Jeremy
Why don't you try a decent quality attenuator first?
The PGA2311 volume control may be very good indeed, but why introduce another electronics stage into the chain, which will undoubtedly have some kind of sonic signature of it's own if you do not need to.I am not saying that attenuators do not have any kind of sonic signature, of course they will have a minor signature due to the type of resistors used. A good attenuator may also have 5 times the bandwidth too.
Alex
Why don't you try a decent quality attenuator first?
The PGA2311 volume control may be very good indeed, but why introduce another electronics stage into the chain, which will undoubtedly have some kind of sonic signature of it's own if you do not need to.I am not saying that attenuators do not have any kind of sonic signature, of course they will have a minor signature due to the type of resistors used. A good attenuator may also have 5 times the bandwidth too.
Alex
I'm going to try an attenuator too - I'm trying everything out to learn how it works and see how it sounds! 🙂
jp_howard said:
Whilst this regulator will give you some predetermined voltage it is textbook crude , no filter on the noisy zener and no output filteringb best used on a constant load, if you want something sophisticated an to get your teeth into try this, the previous version with a 317/337 pre reg works very well, but not as good as a shunt regulator.
http://www.pinkfishmedia.net/forum/showthread.php?t=61445
Alfred
Actually I've had a slight change of plan today - I've decided to put the PGA2311 into the chipamp I plan to build, and an attenuator into the SCULD. I want to have 2 amps that are as different as possible - one small digitally controlled chip-amp, and one big discrete amp. It'll be interesting to compare the two.
I'm very keen to do some of my own measurements at home, both of this amp and the chipamp that I plan to build. I've found some good software (Visual Analyzer), and I've gotten hold of a suitable sound card (Asus Xonar Essence STX: input THD of 0.0002% and output THD of 0.0003%; 192kHz/24bit).
I have bought the parts to create a dummy load, as described in http://www.k4eaa.com/dummy.html (except of course my dummy load will be 8Ohm rather than 50Ohm).
So the plan is to feed the amp's signal to the dummy load, and then on to the sound card via some kind of buffer or voltage divider. I've seen some suggestion that just a voltage divider is adequate, but most approaches I've seen use an opamp, such as this one (from http://xoscope.sourceforge.net/hardware/hardware.html):
I'd rather not use an opamp if possible, since I believe it'll add to the THD I'm trying to measure. Since I only want to attenuate the input signal, not amplify it, is the opamp necessary/useful?
Am I going about this the right way? Or should I be doing something else entirely, such as connecting directly to the signal before it hits the output devices?
I have bought the parts to create a dummy load, as described in http://www.k4eaa.com/dummy.html (except of course my dummy load will be 8Ohm rather than 50Ohm).
So the plan is to feed the amp's signal to the dummy load, and then on to the sound card via some kind of buffer or voltage divider. I've seen some suggestion that just a voltage divider is adequate, but most approaches I've seen use an opamp, such as this one (from http://xoscope.sourceforge.net/hardware/hardware.html):
I'd rather not use an opamp if possible, since I believe it'll add to the THD I'm trying to measure. Since I only want to attenuate the input signal, not amplify it, is the opamp necessary/useful?
Am I going about this the right way? Or should I be doing something else entirely, such as connecting directly to the signal before it hits the output devices?
jp_howard said:I'm very keen to do some of my own measurements at home, both of this amp and the chipamp that I plan to build. I've found some good software (Visual Analyzer), and I've gotten hold of a suitable sound card (Asus Xonar Essence STX: input THD of 0.0002% and output THD of 0.0003%; 192kHz/24bit).
I have bought the parts to create a dummy load, as described in http://www.k4eaa.com/dummy.html (except of course my dummy load will be 8Ohm rather than 50Ohm).
So the plan is to feed the amp's signal to the dummy load, and then on to the sound card via some kind of buffer or voltage divider. I've seen some suggestion that just a voltage divider is adequate, but most approaches I've seen use an opamp, such as this one (from http://xoscope.sourceforge.net/hardware/hardware.html):
I'd rather not use an opamp if possible, since I believe it'll add to the THD I'm trying to measure. Since I only want to attenuate the input signal, not amplify it, is the opamp necessary/useful?
Am I going about this the right way? Or should I be doing something else entirely, such as connecting directly to the signal before it hits the output devices?
Hello Jeremy
Why dont you tell everybody what the imputus to do measurements on your amplifier was, and report on the measurement results that you did at my place last Saturday (2/5/09) .
If you cant remember I can refresh your memory.
Regard
Arthur
Sure - I didn't want to do so without your permission; thanks for letting me pass on the info you provided.PHEONIX said:
As Arthur mentions, we did measurements on the weekend of harmonic distortion using his very nice Audio Precision audio analyzer. The impetus was that Arthur had previously tested another person's SCULD, and was unable to match the figures reported in Silicon Chip. He had found THD at 1kHz in line with SC's article (0.0008%), but substantially higher THD at higher frequencies that the article reported.
These results were replicated on my amp. I don't have the measurements written down, but I'm sure Arthur can provide the actual data. The article reports 0.004% into an 8Ohm load at 10k - IIRC our observation was at least an order of magnitude higher than that.
Whilst clearly measurements like this are not a sufficient test to decide that an amp is any good, I did find it an extremely useful exercise to confirm that it is working correctly. For instance, the measurements picked up that one of the two channels had some 50Hz noise, due to a grounding problem (the 2 star grounds are connected through the chassis, which Arthur explained is a Bad Idea - also, the RCA sockets are not currently insulated from the chassis, which is another Bad Idea).
The experience has made me very enthusiastic about being able to do my own testing at home, hence my request above for help in getting measurements from my sound card. So if anyone has any suggestions on the various questions above, please let me know! 🙂
Jeremy
Are you able to confirm that in your remaining bits and pieces,that all supplied transistors in the purchased kits, other than the 2SA970, which must be Toshiba, are from Philips, (NXP Semiconductors) ON Semiconductor, and ST Microelectronics ?
Alex
P.S. Just curious.
Are you able to confirm that in your remaining bits and pieces,that all supplied transistors in the purchased kits, other than the 2SA970, which must be Toshiba, are from Philips, (NXP Semiconductors) ON Semiconductor, and ST Microelectronics ?
Alex
P.S. Just curious.
Certainly the MJE's and NJL's are from ON, since they're clearly labelled on the cases. Dunno about the other trannies however - I can't see any clear markings on them to indicate the manufacturer. I put in bc550c's that I bought separately in order to get their higher HFE, BTW. I should also add that my amp isn't identical to the article's design, since it has trimpots for biasing and current mirror matching, 0R22's instead of 0R1's, etc...
Jeremy
Those changes are more likely to improve performance at 10KHZ.
Of more interest would be the manufacturers of the BC639 and BF469 and BF470.
Alex
Those changes are more likely to improve performance at 10KHZ.
Of more interest would be the manufacturers of the BC639 and BF469 and BF470.
Alex
jp_howard said:
The measured thd result can depend significantly on the measurement bandwidth. SC usually quote a 80KHz measurement BW which may be considerably less than you used - unless you have changed the input filter from the original values.
I would break up the dummy load into a 0R5 and 7R5 in series with the 0R5 near 0V and tap off the voltage between the resistors. The low output impedance will ensure the lowest possible THD from your sound card
BR
VHF man said:
Hello BR
Indeed 80kHz BW was used but I also limited the bottom end to 400Hz to eliminate power supply issues (100 Hz and some harmonics) using a built in switchable high pass filter . If I removed this high pass filter the analyzer is flat down to 10Hz and of coarse the THD+N gets worse.
Regards
Arthur
PHEONIX said:
Hello Arthur
Sorry, the BR was simply an abbreviation (= best regards).
My actual name is David.
Thanks for clarifying the measurements - although it sheds no light on why the discrepancy.
Perhaps the quiescent current was not optimised?
I think you would need to look at the actual THD spectra to work out the most likely origins of the distortion.
Regards,
David
Thanks for the advice. Can I just make sure I understand what you're recommending? What you're suggesting is to use the dummy load as a voltage divider, and use that voltage to directly connect to the sound card, without using any other circuit or components in between?VHF man said:
If so, perhaps even better would be to put a few 0R1's in series with the main dummy load, and use a dial to allow the effective gain of the voltage divider to be adjusted as required, by adding/removing those 0R1's. 🙂
Would you expect much/any noticeable impact on distortion of such an approach?
jp_howard said:
Your understanding of what I was trying to say is 100% correct. My recommendation to use the 7.5 + 0.5 was simply to retain the 8 ohm load with an approximated (best guess) attenuation, but you could change total R by 20% without any significant change in THD. Yes, you could use a few resistors in a ladder and tap from the most optimum point after trial / error but I would not recommend the use of any in line switching because even a small amount of contact resistance can increase measured THD. Your typical Jaycar selector switch would be terrible at these currents.
If you know what the voltage gain is ((R1+R2)/R1) then you can calculate the dummy load resistors for the correct attenuation so that the amplifier gain is 1 - which should be what you want.
Hope that makes sense.
That makes perfect sense, and has the benefit that it's even within the scope of my limited capabilities. It'll be a good little project for me to start on tonight whilst Evelyn is off playing tennis. I've even got a spare heatsink I can stick the resistors on to in order to allow a higher power for my dummy load (which I think is a bit easier to modify later than the mineral oil bath suggested in the link I provided earlier).
I don't think we took a screen-shot of the actual spectra, but IIRC the level of each harmonic decreased fairly linearly from the 2nd onwards. There wasn't any other substantive distortion.VHF man said:
jp_howard said:
That's a good sign. The ratio of 2nd to 3rd harmonic can be a pointer to how well the LTP/CM or output stage are cancelling even order distortion (symmetry). If you're not seeing high order components then it's a good bet that the quiescent current is optimum or close to it. The silicon chip guys probably selected the LTP and output devices for close (HFE and VBE) matching. This can make THD specs look better (less even order) but depending on your persuation it may or may not improve the subjective result. What is unanimously considered a good thing is the low content of high order harmonics.
FWIW I generally prefer lowest THD possible.
- Status
- Not open for further replies.