This may be in the wrong forum, but my goal is to be able to see THD and IM distortion before and after I change components or values in power amps.
I have a few software packages available for my PC which has a good isolated external sound card, but in order for the results to have value I need to build a resistor dividing attenuation network of some sort connected to an appropriate load for my testing.
Since I am trying to measure THD, etc; I want a quiet solution and also don't want to accidentally send +-80 volts into my sound board. I realize that the resistors would need to be sized appropriately for the voltage range I use, but my question is more of what are appropriate types of resistors, and what are good starting sizes. It seems like high tolerance metal film resistors maybe? By size I mean, if I want a 10:1 ratio would it be better to use 10 ohm to ground and 100ohm to source? or 1kohm to ground and 10kohm to source?
My test load will be 8,4,2 ohms. How much additional load do can I put on the circuit to get good signal, but not effect the load the amp sees significantly enough for the output to change. .1 ohm, .01ohm, .001 ohm? I would also prefer to not have to use huge resistors for the dividing network, and would prefer for them not to get smoking hot.
My end goal would be to have a +-1.8 volt signal that is an exact clone of, say, a +-20 volt signal. I need the signal to be as clean as possible to measure.
I have a few software packages available for my PC which has a good isolated external sound card, but in order for the results to have value I need to build a resistor dividing attenuation network of some sort connected to an appropriate load for my testing.
Since I am trying to measure THD, etc; I want a quiet solution and also don't want to accidentally send +-80 volts into my sound board. I realize that the resistors would need to be sized appropriately for the voltage range I use, but my question is more of what are appropriate types of resistors, and what are good starting sizes. It seems like high tolerance metal film resistors maybe? By size I mean, if I want a 10:1 ratio would it be better to use 10 ohm to ground and 100ohm to source? or 1kohm to ground and 10kohm to source?
My test load will be 8,4,2 ohms. How much additional load do can I put on the circuit to get good signal, but not effect the load the amp sees significantly enough for the output to change. .1 ohm, .01ohm, .001 ohm? I would also prefer to not have to use huge resistors for the dividing network, and would prefer for them not to get smoking hot.
My end goal would be to have a +-1.8 volt signal that is an exact clone of, say, a +-20 volt signal. I need the signal to be as clean as possible to measure.
If you're going for a clean signal, resistors are the way to go.
It might be of interest to test using an actual speaker as well as the dummy load to test the amp's performance into a realistic load.
It should help to know the input impedance of your card. If we say it's 10k, then we can just hook it in series with a 90k resistor to give 1/10 division.
Using a 90 ohm and 10 ohm resistor combination for 1/10 shouldn't throw off the amp loading enough to cause any difference. In any case, unless you know your card's impedance, you might not get perfect 1/10 division (as long as the voltage is in your card's rage it shouldn't be a problem).
If you test using an actual speaker I advise against taking the output voltage from a low value resistor in series with it. This is because even though the amp is controlling the voltage, the speaker responds differently to every signal and your output will be more of a test of the speaker than of the amp.
If you are using a trusted 8 ohm dummy load you can put a .1ohm resistor in series and take your input from across that resistor. Then you will have approximately 1V for every 80V, so your card won't risk being smoked. For a 4 ohm load half this value and for a 2 ohm load 1/4.
- keantoken
It might be of interest to test using an actual speaker as well as the dummy load to test the amp's performance into a realistic load.
It should help to know the input impedance of your card. If we say it's 10k, then we can just hook it in series with a 90k resistor to give 1/10 division.
Using a 90 ohm and 10 ohm resistor combination for 1/10 shouldn't throw off the amp loading enough to cause any difference. In any case, unless you know your card's impedance, you might not get perfect 1/10 division (as long as the voltage is in your card's rage it shouldn't be a problem).
If you test using an actual speaker I advise against taking the output voltage from a low value resistor in series with it. This is because even though the amp is controlling the voltage, the speaker responds differently to every signal and your output will be more of a test of the speaker than of the amp.
If you are using a trusted 8 ohm dummy load you can put a .1ohm resistor in series and take your input from across that resistor. Then you will have approximately 1V for every 80V, so your card won't risk being smoked. For a 4 ohm load half this value and for a 2 ohm load 1/4.
- keantoken
its unlikely any soundcard line level front end is much quieter than a 1K Ohm resistor so 100 - 1K is likely a fine target range for the lower divider resistor - using a lower impedance divider only requires inconveniently large size to dissipate the power without adding to measurement accuracy – the same consideration applies to the amplifier’s own global feedback network as well
metal film at <10% of their power rating are good to ppm for audio frequencies much higher than their thermal time constant
an amusing hack is to use a string of identical resistors for the divider – they all see the same Vdrop, and power so if the thermal environments are kept closely similar any common distortion mechanism like TC or VC will ratio out
metal film at <10% of their power rating are good to ppm for audio frequencies much higher than their thermal time constant
an amusing hack is to use a string of identical resistors for the divider – they all see the same Vdrop, and power so if the thermal environments are kept closely similar any common distortion mechanism like TC or VC will ratio out
Last edited:
The disadvantage of a fixed divider is sacrificing your card's dynamic range for measurements at any level but one. Ideally you want to drive the sound card as close to full permissible level as possible. Seriously, use a pot. I regularly take 96/24 measurements with a generic plastic A/B pot, a 10-turn wire wound will be massive performance overkill.
I use the silicon chip sound card preamp, though it does have the disadvantage that it lowers your measuring resolution somewhat, in that it of course has it's own noise and distortion profile.
One thing that is important though is to make sure you measure your sound card in loop mode (ie line out directly connected to line in) so that you can get a baseline measurement of what the noise and distortion performance of your soundcard is!
Tony.
One thing that is important though is to make sure you measure your sound card in loop mode (ie line out directly connected to line in) so that you can get a baseline measurement of what the noise and distortion performance of your soundcard is!
Tony.
Thanks for the suggestions.
For the initial setup I will definately test with some old expendable sound cards first.
Ultimately I will either end up using an M-audio 1010, which is my regular sound card. Or if the results are similar I will use either my USB Turtle beach, or USB creative Soundblaster. I am not worried about them getting fried. If the results are much better I will use the M-audio, but I don't want to break that one so I may only use it for fine tweaking.
Who knows - I may end up giving up on the software if it isn't great.
For the initial setup I will definately test with some old expendable sound cards first.
Ultimately I will either end up using an M-audio 1010, which is my regular sound card. Or if the results are similar I will use either my USB Turtle beach, or USB creative Soundblaster. I am not worried about them getting fried. If the results are much better I will use the M-audio, but I don't want to break that one so I may only use it for fine tweaking.
Who knows - I may end up giving up on the software if it isn't great.
i use Spectrum Lab. it has a LOT of flexibility and features.
http://freenet-homepage.de/dl4yhf/spectra1.html
it can use sound cards with up to 196k sample rates
http://freenet-homepage.de/dl4yhf/spectra1.html
it can use sound cards with up to 196k sample rates
If your divider network has high impedance, is it possible that the double zeners would cause significant distortion at HF because of Varactor-type capacitance, or am I overdoing it?
Many zeners have (according to my SPICE models) Cj of 100-200pF.
Maybe I'm overreacting to details. I would use strings of 1N4148 in both directions, for less than 4pF of capacitance, then there's no chance for worry unless you're operating near overdrive.
- keantoken
Many zeners have (according to my SPICE models) Cj of 100-200pF.
Maybe I'm overreacting to details. I would use strings of 1N4148 in both directions, for less than 4pF of capacitance, then there's no chance for worry unless you're operating near overdrive.
- keantoken
you could always add compensation capacitance to the divider if the junction capacitance of the zener causes any problems. using a sound card, you won't have enough bandwidth for 100pf to be much of a problem. you will need at least a 196khz sampling sound card to do THD20k measurements anyway.
let's assume a 100:1 10k voltage divider (9.9k+100 ohms) and a zener with 100pf junction capacitance. the turnover frequency for 9.9k and 100pf is 160khz which is past the nyquist limit of even a 192k sound card (which would be 96khz). the capacitance gets smaller with voltage, but even then, you're only putting 1V across the zeners. the capacitance change is only going to be a few pf. the cable capacitance going into the sound card will swamp the zener capacitance changes. Xc at 20khz of 100pf is 80k, which is swamped by the 100 ohm resistor
Last edited:
hi all,
i to am thinking about exactly the same thing, i found this on RMAA forum a while back, idea looks good. needs someone to put it into a practical form.
the part of the circuit 'voltage controled dual clipper' looks to be the basis of a great idea, needs refinment though particular the reference voltages.
this is on my to do list about No 347562
anyway hope this is of interest
i to am thinking about exactly the same thing, i found this on RMAA forum a while back, idea looks good. needs someone to put it into a practical form.
the part of the circuit 'voltage controled dual clipper' looks to be the basis of a great idea, needs refinment though particular the reference voltages.
this is on my to do list about No 347562
anyway hope this is of interest
Attachments
Last edited:
Looks good. But would it work for very low THD measurements?
And could it be made simpler with respect to V4 and V5 references? My impression is that for something like this to gain interest, it would have to be virtually transparent (IE unmeasurable or extremely low THD), and it would also have to be simple (that LT1920 might scare someone away).
In any case I think most just use the methods already given in this thread, so they probably don't feel the need to design/build a whole circuit to do it for them. This kind of thing would be sold as a module included with analyzing software and a data acquisition card. Alone it might not generate much interest.
- keantoken
And could it be made simpler with respect to V4 and V5 references? My impression is that for something like this to gain interest, it would have to be virtually transparent (IE unmeasurable or extremely low THD), and it would also have to be simple (that LT1920 might scare someone away).
In any case I think most just use the methods already given in this thread, so they probably don't feel the need to design/build a whole circuit to do it for them. This kind of thing would be sold as a module included with analyzing software and a data acquisition card. Alone it might not generate much interest.
- keantoken
Some ideas here too http://www.electronic-engineering.ch/radiocontrol/circuits/osziconn/connect.html
Or this
Or this
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
nice, except almost all the resistor values shown is going to require high frequency compensation caps. i used a similar device with a dummy load to go into an oscope, and without compensation, flat top peaks were distorted by the flat tops being angled (at a 45 degree angle).
but that's pretty much what i had in mind as far as the zeners. all you really need for amplifiers is a x100, x50, x20 and x10, and maybe x5 and x2. anything beyond that isn't really neccesary because for most amplifiers you won't see signals beyond +/-100V. just keep in mind that a divider using a 100 ohm resistor as the base of the string is only going to be for use with dummy loads (because with dummy loads you can get away with having a low impedance voltage divider). for other uses such as using the sound card as general purpose oscope or spectrum analyzer, you will want to use a much higher resistance string such as the one in the pic above. with something like that you will find you need compensation caps. i usually use 30, 50 or 100pf trimmers. adjust the compensation by inputting a square wave of the appropriate voltage into the attenuator and adjusting the trimmer for flat tops on the waveform. the compensation caps should only have to be set once unless you change cables between the attenuator and sound card.
but that's pretty much what i had in mind as far as the zeners. all you really need for amplifiers is a x100, x50, x20 and x10, and maybe x5 and x2. anything beyond that isn't really neccesary because for most amplifiers you won't see signals beyond +/-100V. just keep in mind that a divider using a 100 ohm resistor as the base of the string is only going to be for use with dummy loads (because with dummy loads you can get away with having a low impedance voltage divider). for other uses such as using the sound card as general purpose oscope or spectrum analyzer, you will want to use a much higher resistance string such as the one in the pic above. with something like that you will find you need compensation caps. i usually use 30, 50 or 100pf trimmers. adjust the compensation by inputting a square wave of the appropriate voltage into the attenuator and adjusting the trimmer for flat tops on the waveform. the compensation caps should only have to be set once unless you change cables between the attenuator and sound card.
Last edited:
So I set up two different divider networks - one with a pot, which works very well, and one with fixed resistors. Except that with each network I am getting a resonance peak of some sort. With one network it is at 5K, 70db below 0, the other about 400hz, also about 70db below 0.
I assume I need to put a tiny cap across something? Unfortunately I don't have any nf or pf caps here.
I also tried just using a 1x/10x oscilloscope probe grounded and connected to the input of my sound card - it also works fairly well and has a switch! I have not figured out what the actual ratio is with it, it's load is in the megaohm range.
Question 5 - trying to measure noise at low output, can I hook the soundcard input directly to the amp output (carefully and with protection diodes)? I want to look at the amp output with <1 volt output. or should I still use resistors to drop the level?
I assume I need to put a tiny cap across something? Unfortunately I don't have any nf or pf caps here.
I also tried just using a 1x/10x oscilloscope probe grounded and connected to the input of my sound card - it also works fairly well and has a switch! I have not figured out what the actual ratio is with it, it's load is in the megaohm range.
Question 5 - trying to measure noise at low output, can I hook the soundcard input directly to the amp output (carefully and with protection diodes)? I want to look at the amp output with <1 volt output. or should I still use resistors to drop the level?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.