Anyone seen something like this before?:
I picked it up the other day before getting the tip about using the dummy load as a voltage divider, with plans to use it to drop the voltage to the sound card. Looks like a simple combination of a 5W resistor, little transformer, and a trimpot. Might be worth trying as well? Or likely to add more distortion than the voltage divider approach?
I picked it up the other day before getting the tip about using the dummy load as a voltage divider, with plans to use it to drop the voltage to the sound card. Looks like a simple combination of a 5W resistor, little transformer, and a trimpot. Might be worth trying as well? Or likely to add more distortion than the voltage divider approach?
Dummy load resistors attached to heatsink, and low THD sound card ready to install: 🙂
Nichicon Fine Gold caps, and replacable opamps - not stuff you see in a PC sound card every day... 😎
Nichicon Fine Gold caps, and replacable opamps - not stuff you see in a PC sound card every day... 😎
Hi all,
One thing to watch for when doing measurements down at these low levels is to know the noise floor of (in this case) the Audio Precision audio analyzer. 0.00 anythink can be hard to measure.
Terry
One thing to watch for when doing measurements down at these low levels is to know the noise floor of (in this case) the Audio Precision audio analyzer. 0.00 anythink can be hard to measure.
Terry
jp_howard said:
>0.04% THD at only 10kHz indicates that you guys have seriously screwed the pooch with this amplifier somewhere along the line; it is not a case of fraud on behalf of SC.
Was it measured to 20kHz? If so how bad was it there? Was it on the verge of clipping?
0.22 ohm vs. 0.1 ohm emitter resistors or a 80kHz vs. 200kHz measurement bandwidth would not account for this discrepancy.
Most likely a grounding issue / layout issue. How many fudges and tweaks have been applied to the PCB and components?
G.Kleinschmidt said:
>0.04% THD at only 10kHz indicates that you guys have seriously screwed the pooch with this amplifier somewhere along the line; it is not a case of fraud on behalf of SC.
Was it measured to 20kHz? If so how bad was it there? Was it on the verge of clipping?
0.22 ohm vs. 0.1 ohm emitter resistors or a 80kHz vs. 200kHz measurement bandwidth would not account for this discrepancy.
Most likely a grounding issue / layout issue. How many fudges and tweaks have been applied to the PCB and components?
Hello Glenn
I measured it as per SC article.
This is a blameless amp with new Onsemi thermotrak parts as you know. I would like to know what somebody elses independent measurements of this blameless amp are . Do you own or have access to a blameless amp and can you measure it . If you can lets make the bench mark power and load 100Watts into 8R and you tell me what THD+N you measure at 1Khz, 10KHz, and 20KHz (80Khz BW, its also a good idea to high pass the bottom end at 400Hz ). If the power benchmark is too high you you just tell me what you would like it to be.
It could be I have a problem with the AP.
Regards
Arthur
Glenn
I believe that Jeremy's build is different in that it uses 2 separate toroidal transformers. Also, the input RCAs were not insulated from the case, and there would also have been other earthing layout differences due to "dual mono" construction.
SandyK
I believe that Jeremy's build is different in that it uses 2 separate toroidal transformers. Also, the input RCAs were not insulated from the case, and there would also have been other earthing layout differences due to "dual mono" construction.
SandyK
I'm sure that OPS is underbiased. That is the reason why it measure bad at 10Khz.
Please try to measure again with 180mA through 0r1 Re
Please try to measure again with 180mA through 0r1 Re
Alex, Arthur's other SCULD he tested doesn't have those same issues, but has the same measurements. Once I've got my testing rig finished (assuming I can get it to work OK), I hope to test the impact of using just one toroidal, insulating the RCA's, and other similar changes. I want to get the testing setup working first however, so I can measure the impact of each change.
Roender: SC's article says that target measurement is 7-10mV across the 0R1's. Arthur's other amp matches this - my amp was tested with 24mV across 0R22's. Although you may feel that this is too low, this is the SC spec, so I guess that's what their measurements were based on. (Although I'm not sure how - the circuit as shown in the article has <2mV across the 0R1's, since it doesn't include a trimpot!)
Roender: SC's article says that target measurement is 7-10mV across the 0R1's. Arthur's other amp matches this - my amp was tested with 24mV across 0R22's. Although you may feel that this is too low, this is the SC spec, so I guess that's what their measurements were based on. (Although I'm not sure how - the circuit as shown in the article has <2mV across the 0R1's, since it doesn't include a trimpot!)
Regardless - the issue is, why the discrepancies in THD and mV measurements between the SC article and independent testing. Based on these two issues, it seems that the SC authors may have had a somewhat different configuration to what's shown in the article. That's just one hypothesis - I don't really know what's happening here, but I am curious to track it down.roender said:
The noise floor of the AP is shown here: http://ap.com/products/2700/analog_performance . -112dB.pheonix358 said:
When I do this for my test rig, is it sufficient to simply use a loop back cable and test the noise of that? I guess better would be to put the dummy load in there as well, since that'll add to the noise floor...
BTW, it turns out that my new sound card line in can handle 5.6Vpp, so I decided on just a 0.1x attenuation in my voltage divider (in the dummy load). I've tested it now, and it even works! 😀 Provides (roughly) the correct load, and drops the voltage appropriately. Just waiting for the glue to dry before I do some amp testing - should have time for that tomorrow with a bit of luck.
PHEONIX said:
I second Arthur's findings. I was there when he done them as well as test my Yamaha M-70 which had an order of magnitude lower distortion because of its error correction circuit and high fT power transistors.
Either his AP is not functioning properly or there is an inherent distortion mechanism at play. But his sound card and Spectralab software also confirms what he is seeing on his AP 😉
Me thinks it is switching distortion due to minority carrier storage and common mode conduction mechanism.
But like others I'm curious to know how SC and Douglas Self come up with such optimistic measurements when they both use the same test equipment as Arthur 😕
Well I've got my Asus Xonar set up now. It comes with an Audio Precision test report that shows THD+N of 0.0003% on line out and 0.0002% on line in. I've tested a 3 meter shielded RCA loopback cable and got 0.0007% THD+N, and 0.0002% THD.
I'm quite amazed such low noise can be obtained in a noisy PC case! I did find that I get the best results when my PC isn't busy with other things - for instance when the TV tuner was busy recording the footy last night, THD went up to 0.0005% from 0.0002%. It's also sensitive to cable - a 1.5 meter unshielded RCA cable had THD of 0.0007%.
Anyways... it looks like I've got a low enough noise floor now to test most amps, including the SCULD - assuming that the voltage divider doesn't add too much noise. Next step it to test that.
BTW I replaced the lm4562 DIP in the sound card with the 'metal cap' version, and thermally bonded it to the metal shield on the sound card, to use as a heat sink. The thermal properties of this version of the chip seem better than the DIP version.
I'm quite amazed such low noise can be obtained in a noisy PC case! I did find that I get the best results when my PC isn't busy with other things - for instance when the TV tuner was busy recording the footy last night, THD went up to 0.0005% from 0.0002%. It's also sensitive to cable - a 1.5 meter unshielded RCA cable had THD of 0.0007%.
Anyways... it looks like I've got a low enough noise floor now to test most amps, including the SCULD - assuming that the voltage divider doesn't add too much noise. Next step it to test that.
BTW I replaced the lm4562 DIP in the sound card with the 'metal cap' version, and thermally bonded it to the metal shield on the sound card, to use as a heat sink. The thermal properties of this version of the chip seem better than the DIP version.
Jeremy
I hope the LM4562HA case is insulated from the shield, as the case of the LM4562HA is connected to the -VE supply rail. Not so important in a single supply rail setup, but not a good idea .
Alex
I hope the LM4562HA case is insulated from the shield, as the case of the LM4562HA is connected to the -VE supply rail. Not so important in a single supply rail setup, but not a good idea .
Alex
Thanks Alex - I didn't know that. I just checked, and the heat transfer tape I used isn't conductive.
It's taken me a while to get meaningful results with the dummy load / voltage divider. The main issue was that my first attempt was pretty sloppy, and it ended up introducing a lot of distortion. The other issue was that actually getting a measurement of the THD of the dummy load, without using an amp, was tricky, since the output of the voltage divider was too low to be very useful.
To fix the first issue, I put it in a shielded box, and used decent connectors and internal cabling.
To fix the second issue, I created a couple of cables that allowed me to put the 8Ohm dummy load (without the voltage divider) in the middle of a loopback. I then tested the THD of this across a few frequencies from 1k to 10k, at a few dB below clipping:
The result should be good enough for the simple testing I want to do of the SCULD - the issues I'm trying to resolve that I saw with Arthur (high distortion and high frequencies; high distortion in one of the 2 channels) are at levels an order of magnitude higher THD than the noise floor of the sound card plus dummy load combined.
BTW, I found that different programs gave surprisingly different spectra, and slightly different THD measurements, even when using all the same settings (windowing, FFT params, samples, etc). Here's the spectra/THD for Visual Analyzer, ARTA, and RMAA, respectively:
VA:
ARTA:
RMAA:
(Note that windowing and FFT params can't be modified in the version of RMAA I'm using, so it's not quite comparable to the other two.)
VA and ARTA show pretty much the same thing, but the harmonics at 1333Hz, 1666Hz, etc, don't appear in RMAA's graph at all. I'm not at all sure what that means... 😕
To fix the first issue, I put it in a shielded box, and used decent connectors and internal cabling.
To fix the second issue, I created a couple of cables that allowed me to put the 8Ohm dummy load (without the voltage divider) in the middle of a loopback. I then tested the THD of this across a few frequencies from 1k to 10k, at a few dB below clipping:
The result should be good enough for the simple testing I want to do of the SCULD - the issues I'm trying to resolve that I saw with Arthur (high distortion and high frequencies; high distortion in one of the 2 channels) are at levels an order of magnitude higher THD than the noise floor of the sound card plus dummy load combined.
BTW, I found that different programs gave surprisingly different spectra, and slightly different THD measurements, even when using all the same settings (windowing, FFT params, samples, etc). Here's the spectra/THD for Visual Analyzer, ARTA, and RMAA, respectively:
VA:
ARTA:
RMAA:
(Note that windowing and FFT params can't be modified in the version of RMAA I'm using, so it's not quite comparable to the other two.)
VA and ARTA show pretty much the same thing, but the harmonics at 1333Hz, 1666Hz, etc, don't appear in RMAA's graph at all. I'm not at all sure what that means... 😕
OK, here's the 1st THD analysis chart of my SCULD. It has three series:
You can see that my measurements across the board are not as good as SC, and that one of my channels is worse than the other. As discussed earlier, this is likely to be due to my grounding issues, which I'll be working on shortly. The other amp that Arthur tested showed the same results as SC at 1kHz, so my 1st goal is to get to that point.
Interestingly, for my better channel, the THD at all freqs is a fairly consistent 2.75x worse that SC's measurements. So perhaps in my case the problems at higher freqs have the same cause as the probs at 1kHz - in which case fixing one would fix the other too...
FYI, here's the spectra from Visual Analyzer for one of my channels:
(1kHz)
(7.5kHz - note different x-axis range)
- Silicon Chip's measurements @8Ohm
- My left channel
- My right channel
You can see that my measurements across the board are not as good as SC, and that one of my channels is worse than the other. As discussed earlier, this is likely to be due to my grounding issues, which I'll be working on shortly. The other amp that Arthur tested showed the same results as SC at 1kHz, so my 1st goal is to get to that point.
Interestingly, for my better channel, the THD at all freqs is a fairly consistent 2.75x worse that SC's measurements. So perhaps in my case the problems at higher freqs have the same cause as the probs at 1kHz - in which case fixing one would fix the other too...
FYI, here's the spectra from Visual Analyzer for one of my channels:
(1kHz)
(7.5kHz - note different x-axis range)
I don't know, but some of those spikes look like they are external noise, ie, computer noise. BTW, your equipment noise floor needs to be considerably lower than the equipment being tested. With amps like these it does get tricky. When you are next Taking a measurment and one of those spikes appear, try looking at your computer cables, orientation of amp vs computer, change something and try again. I think those spikes are external to rather than an issue with the amp. Testing can be fun.
Terry
Terry
Thanks for the comments Terry. I posted pics and details of my noise floor in the previous post. At 1kHz, my noise floor is 50% of the total THD, and at 10kHz, it is only 10% of the total THD. I have already removed the noise floor from the THD comparison chart, BTW. Since it is quite consistent, and is an order of magnitude lower than the amp's THD at high frequencies, it's not going to get in the way of the analysis I want to do of my amp.
The noise in the noise floor isn't from the sound card, as I showed a few days ago:
That's the sound card on it's own. THD is only 0.00017%! That THD is fairly constant up to 10kHz. The noise floor I'm getting is from the dummy load and voltage divider, not the sound card. I've already got it down 10x lower than the crappy job I did on my first attempt, but there's still plenty of room to improve (e.g. using something like http://australia.rs-online.com/web/search/searchBrowseAction.html?method=getProduct&R=2251137 ).
The noise in the noise floor isn't from the sound card, as I showed a few days ago:
That's the sound card on it's own. THD is only 0.00017%! That THD is fairly constant up to 10kHz. The noise floor I'm getting is from the dummy load and voltage divider, not the sound card. I've already got it down 10x lower than the crappy job I did on my first attempt, but there's still plenty of room to improve (e.g. using something like http://australia.rs-online.com/web/search/searchBrowseAction.html?method=getProduct&R=2251137 ).
Ok, you seem to have it under control, I am wondering what that great thin phallic symbol is at 1K. 🙂. If that was a speaker (my area of expertise) you would have a problem.
Just curious
Terry
Just curious
Terry
- Status
- Not open for further replies.