Just to be really clear, the junction temp is the internal temp of the regulator right at the little hunk of silicon that's doing all the heavy lifting. It's the hottest part of the regulator. The "case temp" is the outer heat sink contact area of the regulator just like sofaspud suggests. The regulators in the O2 are rated for a junction temp rise of 40 - 60 degrees C per watt of power they have to dissipate when used without a heatsink.
The "rise" means above whatever the ambient temp is. With the board sitting out in free air, that temp is typically only about 5, or worst case, 10 degrees C above the room temp so 25 or 30C. Even with the board in the case, it doesn't change that much as there's not much else in the case to produce significant heat.
I have a remote temp probe and have measured the actual temps in an out of the case with the amp working far harder that it can into HD650s. The draw of the amp, and hence load on the regulators, is directly proportional to the headphone impedance and output level. The actual measurements, using the WAU16-400 that gets the regulators quite a bit warmer than a 12 VAC transformer, are consistent with what I posted two pages ago.
CSA's claim of 77 C on the outside of the regulators would mean a junction temp well over 80C. Using his own current draw numbers the regulator tab temp should be under 50 C which is a fairly large difference and not at all "inline" with my numbers. And I've explained why.
The fact is, several here and elsewhere have posted using the HD650's or very similar headphones with nothing getting too warm. Instead of accepting something is obviously wrong with his amp, and providing the requested information to find a solution, CSA seems much more interested in continuing to throw rocks as he's done in dozens of posts on another forum and just did again above.
The "rise" means above whatever the ambient temp is. With the board sitting out in free air, that temp is typically only about 5, or worst case, 10 degrees C above the room temp so 25 or 30C. Even with the board in the case, it doesn't change that much as there's not much else in the case to produce significant heat.
I have a remote temp probe and have measured the actual temps in an out of the case with the amp working far harder that it can into HD650s. The draw of the amp, and hence load on the regulators, is directly proportional to the headphone impedance and output level. The actual measurements, using the WAU16-400 that gets the regulators quite a bit warmer than a 12 VAC transformer, are consistent with what I posted two pages ago.
CSA's claim of 77 C on the outside of the regulators would mean a junction temp well over 80C. Using his own current draw numbers the regulator tab temp should be under 50 C which is a fairly large difference and not at all "inline" with my numbers. And I've explained why.
The fact is, several here and elsewhere have posted using the HD650's or very similar headphones with nothing getting too warm. Instead of accepting something is obviously wrong with his amp, and providing the requested information to find a solution, CSA seems much more interested in continuing to throw rocks as he's done in dozens of posts on another forum and just did again above.
I just read the new diyAudio newsletter and Jan Didden's article on the DSOnano. Luv it! I truly hope this catches on. A diy'er doesn't have to have a $600 Fluke and a $6000 LeCroy. Even some basic modest test equipment will return dividends. Check out Jan's review if you haven't already!
I'm planning to have one of the O2's gain switch positions at 1.6x so I'm thinking that 2.5kΩ (2.49) resisters are in order for R19 and R23. I'm planning to get Vishay RN60D2491FB14 which are rated to 1/4 Watt. I'm also running unity gain in the other position, so no R17 or R21.
Is there anything I need to know that would cause me to change my plan?
Is there anything I need to know that would cause me to change my plan?
Along with personal smear:
CSA"...So I found your response a tad smarmy."
This following on from "Not trying to be an *** ..."
hmmm.
While I love Jan Didden's work, and I applaud affordable test equipment, it's worth pointing out that scopes are only useful for a few measurements of high quality audio gear. They're great for evaluating stability problems, clipping behavior, basic square wave response and severe noise above 96 Khz. But nearly all use 8 bit ADCs and even the "high resolution" versions rarely offer more than 12 bit (ENOB) performance. While the DSOnano uses a 12 bit ADC (built into the MCU) it will be lucky to offer 10 or 11 bit analog performance which is only 60 to 66 dB of dynamic range. CD quality audio has 96 dB of dynamic range and most audio gear aims to at least get close to that if not exceed it. That renders scopes virtually useless for evaluating noise, distortion, crosstalk, etc. which are all typically in the range of -80 dB to -130 dB. The DAC's inherent noise, and the noise in the wideband input amplifier, prevents measuring such low level signals.
That seems OK to me assuming you don't need more gain or at least some more "excess gain" for quiet tracks, etc. See my All About Gain article if you haven't already.
Also, FWIW, there's a false myth circulating (from H-F) you shouldn't leave out the gain resistors in the O2 for 1X gain. It's been suggested to use high value resistors instead (i.e. 10K - 100K) to prevent the input from being an "open circuit". This is pure nonsense. The low impedance 1.5K feedback resistor is what the negative input "sees" when you leave out the gain resistors. Besides increasing the gain every so slightly, there's no difference in performance adding a large value to ground. It's no different than a voltage follower buffer (as the 4556 output stage op amps are configured). Those don't have any resistor to ground on the negative input either.
The 66dB dynamic range figure quoted here is within bandwidth under consideration - half the sample rate of the ADC. However, if the output of the ADC is run into a 4k (say) point FFT there's an effective gain in the dynamic range related to the averaging going on by virtue of acquiring that length of sample. So if you're a DIYer wanting to look for harmonic distortion, the picture is not as bleak as RS suggests because the FFT will allow you to see well below the wideband noise floor - assuming that your MCU has the RAM buffer available to store that length of sample.
As a rough figure 4k points in an FFT gives you around sqrt(2k) gain for each bin in the FFT - in this case more than 30dB. That turns out to be enough to get a reasonable THD measurement, even with only an 11bit ADC. In reality it might be a bit more difficult to 'grab' the spectral components because they're likely to span more than one bin of the FFT output. But the noise 'floor' in each FFT output bin turns out to be below -90dB by virtue of the sharply reduced bandwidth.
I have no idea if this DSOnano has the FFT software installed to do such a thing, or the RAM size to support it. Anyone checked out the detailed specs on the MCU?
Abraxalito brings up a valid point, with a repetitive signal (like a sine wave) you can get significant dynamic expansion through post processing math and/or averaging. But it's only useful for extracting periodic signals from the ~60 dB noise floor. It's useless for extracting random noise and hence many measurements (such as THD+N, noise, dynamic range, etc.). The FFT cannot distinguish between the uncorrelated noise in the device under test, the scope's input amp noise, and the noise in the ADC.
I have half a dozen digital scopes, including a few with high resolution DACs, and none of them do a reasonable job of THD measurement even with their software set up optimally for FFT THD processing. Even a rather lousy $12 USB DAC, like the Syba CM119 I tested, will far outperform all my scopes. And, FWIW, most of my scopes have four and five figure price tags.
So, yeah, if you're trying to measure 0.5% THD from a SET tube amp, it might be possible. But if you're trying to verify a piece of gear is comfortably below the 0.01% (-80 dB) guideline threshold for THD+N, I've yet to see any scope pull that off.
I have half a dozen digital scopes, including a few with high resolution DACs, and none of them do a reasonable job of THD measurement even with their software set up optimally for FFT THD processing. Even a rather lousy $12 USB DAC, like the Syba CM119 I tested, will far outperform all my scopes. And, FWIW, most of my scopes have four and five figure price tags.
So, yeah, if you're trying to measure 0.5% THD from a SET tube amp, it might be possible. But if you're trying to verify a piece of gear is comfortably below the 0.01% (-80 dB) guideline threshold for THD+N, I've yet to see any scope pull that off.
Last edited:
I did forget to mention that the ability to see THD accurately does rather depend on the ADC having better distortion performance than the device under test, which might be a long shot
The original Audio Precision System One Dual Domain's ADC was pretty lousy so wasn't recommended for making THD measurements without the fundamental being first notched. Fortunately ADC technology has advanced considerably since then.The noise floor gain from FFT doesn't depend on having a repetitive signal though - only on the output bins being much narrower bandwidth than the input so the input quantization noise gets spread out thinly.
Suggest if you're in the market for a decent resolution digital scope to check out these guys:
Mixed Signal USB Oscilloscope | Cleverscope
They're the only ones I'm aware of with 14bit ADCs in their front-end.
Sorry if I have been rude, it's really nothing to do with this amp but rather the hype around it. It's a solid amp with good value.
I should point out that I'm far from being an expert on electrical circuits. With that in mind it seems that you're defining the temps purely with the math for power dissipation. I don't see how that is possible. It ignores the IC in use, it ignores the implementation circuit, it basically ignores the heat (aside from giving you a theoretical baseline for heat produced), and it ignores the environment. It's safe to conclude that power dissipation != real world temperature.
The datasheets for MC7812/7912 are vague on the idea of using the recommended circuit or not. I don't know about everyone else, but if a datasheet is vague, I become skeptical. What isn't vague is a picture, and the picture clearly shows a recommended circuit, and this amp doesn't use it. This amp uses the vague suggestion instead. Though it ignores another vague suggestion related to it, (probably because it was vague)...But then you're picking and choosing which vague statements you're going to follow, by George! Which vague statements are closer to the truth??! Solution: Is it really so hard to use the pictured circuit?? It adds one part to each rail, the idea that they even suggested not to is sort of ridiculous when you think about it. I guess if it was a consumer product then cutting corners would be profitable... Anyhow, since this amp doesn't use the best case circuit or parts, but follows the vague suggestion, quoting the datasheet goes out the window.
Good amp, not perfect, it's fixable, just fix it and enjoy I guess.
Thanks for this interesting tangent guys! I expect we're straying more to the objective side of the the objective2 amp that the thread's about ...
If the digital side of most scopes is the weakness, how (functionally) do instruments like the audio precision device mentioned in RS' blog overcome these limitations and still provide the necessary DSP/digitised output? (note I'm new to the audio side of electronics so just asking out of interest. I'm more used to designing gear with MVA type units)
If the digital side of most scopes is the weakness, how (functionally) do instruments like the audio precision device mentioned in RS' blog overcome these limitations and still provide the necessary DSP/digitised output? (note I'm new to the audio side of electronics so just asking out of interest. I'm more used to designing gear with MVA type units)
Its not so much that 'the digital side is the weakness' just that scopes and audio analysers are different horses for different courses. A scope needs wide bandwidth, an audio analyser not. With a 100MHz bandwidth scope, the ADC might typically be sampling at 1Gs/s. With an ADC for audio, perhaps 1/10,000th of that speed. The wider the bandwidth, the greater the noise. Scope screen resolutions are not typically greater than VGA so really there's no point in having more than 10bits of vertical resolution there. Having more would just tend to show up the noise in greater detail
Maybe it's just the name, but I've never been one to migrate automatically towards Fluke. I think I'd rather have something from Agilent like their U1233A. Anybody have input on that? I'm without a DMM at the moment but my neighbor is willing to let me borrow one. I didn't ask anything about the make and model, so no telling if it's any good.
There's a difference of character around here. I'm not good at building circuits but I'm more of the inclination that if something's vague then it is probably very versitile and I get to experimenting in all sorts of ways to figure out what it can do. Think about the first time someone did something out of the ordinary and found it performed better for their design goals... something as simple as running an OpAmp inverted. Someone had to do it first.
All said, I do understand your hesitation. I also have a line I don't cross. What confuses me is in the O2 schematic it looks like the juice is flowing away from every pin on U6. That's not going to prevent me from building it. I don't get how these things work, but if I build it I may get more understanding while fumbling with the tangible parts of all this. And some more reading on the topic is in order.
Last edited:
What a load of.... Please go back to head-fi and stop polluting this thread for the people that have real interest in the O2. Thanks.
If you want to learn basic electronic theory this thread is not the place.
Thanks for that. Would that change to 10 ohms make a more significant difference to battery life than the gain resistor change? And would that change lock the amp in to the low power opamps or would one still be able to plug the normal opamps in just making sure to use higher impedance phones?
The dScope's ADC is a 24 bit device with a greater number of effective bits than nearly any soundcard in existence. As Abraxalito said, it, and the rather critical analog circuits before it (which can handle anything from a few microvolts to 200 volts in 2 dB precision steps), are optimized for use below 100 Khz with extremely low noise. There are some good reasons the dScope, and any similar devices. cost $10,000+. That's especially true when you consider they're made in relatively low volumes.
Every effective bit (Effective Number Of Bits or ENOB) is roughly equal to 6 dB of dynamic range. So a perfect ideal 8 bit DAC with a noiseless amp before it can only manage at best 48 dB of dynamic range, 10 perfect bits is 60 dB, etc. Nobody, to my knowledge, makes an ADC or DAC that delivers true 24 ENOB in a real circuit in the real world. So when you see 32 bit DACs being advertised, some skepticism is in order.
@CSA, you're way off base with your assumptions. Voltage regulators follow simple ohms law calculations with respect to power dissipation. The implementation circuit of the 78xx doesn't matter. It's a 3 terminal device. The dissipation is equal to Pin - Pout and the laws of thermodynamics dictate how hot it will get. The math is completely valid and very well proven. I've explained every aspect of the O2's design, including the power supply and regulators, in detail.
I've also provided real-world measurements to back up my design. Including real world power consumption and temp measurements of the regulators under a variety of conditions. If you have some specific valid criticism of where I've made a mistake in the numbers I've presented, please provide some factual evidence to support your claims?
If you think my amp design is "ridiculous" perhaps you should apply your self professed "far from expert" knowledge at designing, building, perfecting, and testing a better one? Then we can measure them both and may the best amp win. Want to be on YouTube for the challenge?
Every effective bit (Effective Number Of Bits or ENOB) is roughly equal to 6 dB of dynamic range. So a perfect ideal 8 bit DAC with a noiseless amp before it can only manage at best 48 dB of dynamic range, 10 perfect bits is 60 dB, etc. Nobody, to my knowledge, makes an ADC or DAC that delivers true 24 ENOB in a real circuit in the real world. So when you see 32 bit DACs being advertised, some skepticism is in order.
@CSA, you're way off base with your assumptions. Voltage regulators follow simple ohms law calculations with respect to power dissipation. The implementation circuit of the 78xx doesn't matter. It's a 3 terminal device. The dissipation is equal to Pin - Pout and the laws of thermodynamics dictate how hot it will get. The math is completely valid and very well proven. I've explained every aspect of the O2's design, including the power supply and regulators, in detail.
I've also provided real-world measurements to back up my design. Including real world power consumption and temp measurements of the regulators under a variety of conditions. If you have some specific valid criticism of where I've made a mistake in the numbers I've presented, please provide some factual evidence to support your claims?
If you think my amp design is "ridiculous" perhaps you should apply your self professed "far from expert" knowledge at designing, building, perfecting, and testing a better one? Then we can measure them both and may the best amp win. Want to be on YouTube for the challenge?