Bricolo,
let us think practical. To get excellent TIM (SID) and PIM, we need quite high slew rate. High slew rate is quite impossible for low fT. In case we want high fT, we get quite high OLG in the audio band and enough loop gain in audio band as well. If we close FB around pretty linear circuit, we probably win
let us think practical. To get excellent TIM (SID) and PIM, we need quite high slew rate. High slew rate is quite impossible for low fT. In case we want high fT, we get quite high OLG in the audio band and enough loop gain in audio band as well. If we close FB around pretty linear circuit, we probably win
Re: Practical measurement of PIM
The PIM measurements I did were with a coherent IM analyzer. The test signal was the SMPTE 60 & 6000 Hz, 4:1 combination. The analyzer filters out the incoming 60 Hz part of the signal and then a PLL locks onto the 6 kHz "carrier". In-phase (I) and quadrature (Q) detectors then synchronously detect the signal. The output of the I detector is Amplitude Intermodulation (AIM) and the output of the Q detector is Phase Intermodulation (PIM). This instrument is described in detail in my PIM paper on my web site at www.cordellaudio.com. The analyzer as described has a noise floor equivalent to PIM on the order of about 1.5 ns rms. For measurement of my MOSFET power amplifier, I fed the output of the PIM analyzer into a spectrum analyzer to reduce the noise floor by about 30 dB.
You are right, it is very difficult to truly measure the distortion of many modern op amps, and there are no easy solutions to that.
Cheers,
Bob
PMA said:I would be interested how to measure the PIM in the real circuit. The AD797 was mentioned here. There are several opamps even better "to my ears"
The PIM measurements I did were with a coherent IM analyzer. The test signal was the SMPTE 60 & 6000 Hz, 4:1 combination. The analyzer filters out the incoming 60 Hz part of the signal and then a PLL locks onto the 6 kHz "carrier". In-phase (I) and quadrature (Q) detectors then synchronously detect the signal. The output of the I detector is Amplitude Intermodulation (AIM) and the output of the Q detector is Phase Intermodulation (PIM). This instrument is described in detail in my PIM paper on my web site at www.cordellaudio.com. The analyzer as described has a noise floor equivalent to PIM on the order of about 1.5 ns rms. For measurement of my MOSFET power amplifier, I fed the output of the PIM analyzer into a spectrum analyzer to reduce the noise floor by about 30 dB.
You are right, it is very difficult to truly measure the distortion of many modern op amps, and there are no easy solutions to that.
Cheers,
Bob
A closer look to phase modulation
I coded a small program applying phasemodulation of +/-10ns to a single sine wave signal and fft'ing the result. 10ns sounded so small ! But look at the reality what 10ns phase modulation does to a signal...
(I doublechecked, it's really only 10ns)
I find it very interesting that this looks very much like typical harmonic distortions from a poweramp.
Is it possible that this kind of distortion is simply part of normal THD ?
The difference to normal harmonic distortion is that these harmonics are phase shifted 90°. This means 2nd harmonic has phaseshift of +/- 180° instead of +/-90°.
In theory, this distortion can easily be measured with a single sinewave, instead of looking only at the amplitude of the harmonics, you need to look at the phaseshift of these, seperating normal unlinearity and phase modulation.
The next interesting question would be: Can the ear tell the difference between harmonics generated by unlinearity and phase modulation ? (Or, is the phasehift of harmonics audible ?)
If yes, we have a problem, if no, this kind of distortion is just distortion, and will be compensated by enough NFB. (Or, is just part of the normal THD)
But, at least one thing seems for sure, 10ns phase modulation has to be taken serious.
Mike
I coded a small program applying phasemodulation of +/-10ns to a single sine wave signal and fft'ing the result. 10ns sounded so small ! But look at the reality what 10ns phase modulation does to a signal...
(I doublechecked, it's really only 10ns)
I find it very interesting that this looks very much like typical harmonic distortions from a poweramp.
Is it possible that this kind of distortion is simply part of normal THD ?
The difference to normal harmonic distortion is that these harmonics are phase shifted 90°. This means 2nd harmonic has phaseshift of +/- 180° instead of +/-90°.
In theory, this distortion can easily be measured with a single sinewave, instead of looking only at the amplitude of the harmonics, you need to look at the phaseshift of these, seperating normal unlinearity and phase modulation.
The next interesting question would be: Can the ear tell the difference between harmonics generated by unlinearity and phase modulation ? (Or, is the phasehift of harmonics audible ?)
If yes, we have a problem, if no, this kind of distortion is just distortion, and will be compensated by enough NFB. (Or, is just part of the normal THD)
But, at least one thing seems for sure, 10ns phase modulation has to be taken serious.
Mike
Attachments
Bob Cordell said:
I'd change my name to Halcro and make a bunch of money
Just kidding.
Actually, I've been pondering just that question over the last year or so. It is always a good thing to reflect, and its is always a fun thing to take into account new technology or improvements in technology.
I've considered many different input-stage/VAS topologies along the way, but I keep coming back to essentially the one I published. Of course, some evolutionary changes can be made in it.
I would probably aim for a design that does about 150 watts/channel.
I would use faster transistors in the driver and error-correcting stage. I have also built MOSFET power amplifiers with folded emitter followers driving the power MOSFETs, and I like how this topology performs.
I'd still use an output inductor, but it would probably be an air-core torroidal inductor.
I would incorporate a passive-adaptive soft clipping circuit at the input.
I would build it on a 4-layer printed wiring board.
Thanks for your interest.
Bob
Hi Bob,
By any chance can we expect a DIY 100-150 W error correction design from you in the near future? Trust the PC boards will not be Halcro priced!
Folks, I would like to point something out to you:
I personally own a Sony SACD player. It's great! (in features) It plays CD, DVD, and SACD. Wow! How could I ask for more? However, it doesn't use Burr Brown or Analog Devices premium IC's in its audio passband. It uses BA4558 IC's. Look them up folks and compare the schematic to Barrie Gilbert's. They are almost identical.
This is why PIM must be taken into consideration, or else Sony and its competition will not bother to upgrade. Why should they?
I personally own a Sony SACD player. It's great! (in features) It plays CD, DVD, and SACD. Wow! How could I ask for more? However, it doesn't use Burr Brown or Analog Devices premium IC's in its audio passband. It uses BA4558 IC's. Look them up folks and compare the schematic to Barrie Gilbert's. They are almost identical.
This is why PIM must be taken into consideration, or else Sony and its competition will not bother to upgrade. Why should they?
Mike,
Nice work with coding the PIM simulation and showing the spectral result. It looks like 20 ns p-p PIM resulted in a spectral line down about 70 dB, for a distortion magnitude of about 0.03%. Is that right?
Keep in mind that +/- 10 ns is a LOT of PIM. If you look at the measurements I reported in my PIM paper, a 741 op amp at a gain of -10 had PIM of about 4 ns rms. Also, an old 1970's vintage power amp with 20 kHz THD of several tenths of a percent had PIM on the order of 6 ns rms.
At the other extreme, my MOSFET power amp had PIM on the order of 0.1 ns rms, with 0.001% THD-20. Although I'm not trying to suggest a correlation between THD-20 and PIM, one might guess that modern amplifiers with about 0.01% THD-20 would come in at maybe on the order of 1 ns rms PIM. Scaling your results to that would roughly get us PIM-only spectral lines on the order of 0.003%. So maginitude-wise, with admittedly a lot of hand-waiving, the PIM numbers expressed as spectral line amplitudes might be on the order of about 1/3 those of the THD-20.
Cheers,
Bob
Nice work with coding the PIM simulation and showing the spectral result. It looks like 20 ns p-p PIM resulted in a spectral line down about 70 dB, for a distortion magnitude of about 0.03%. Is that right?
Keep in mind that +/- 10 ns is a LOT of PIM. If you look at the measurements I reported in my PIM paper, a 741 op amp at a gain of -10 had PIM of about 4 ns rms. Also, an old 1970's vintage power amp with 20 kHz THD of several tenths of a percent had PIM on the order of 6 ns rms.
At the other extreme, my MOSFET power amp had PIM on the order of 0.1 ns rms, with 0.001% THD-20. Although I'm not trying to suggest a correlation between THD-20 and PIM, one might guess that modern amplifiers with about 0.01% THD-20 would come in at maybe on the order of 1 ns rms PIM. Scaling your results to that would roughly get us PIM-only spectral lines on the order of 0.003%. So maginitude-wise, with admittedly a lot of hand-waiving, the PIM numbers expressed as spectral line amplitudes might be on the order of about 1/3 those of the THD-20.
Cheers,
Bob
Folded emitter followers
Adam,
That's right. The power MOSFET turn-on gate drive is provided by a current source that works against an emitter follower driver of the opposite sex to the power MOSFET. Thus, the driver acts to actively turn off the output transistor, rather than turn it on. If you play your cards right, the driver collectors can be terminated on the output line, so they see low collector voltage and can be fast devices that dissipate little heat. The current source that supplies it all need not be fast transistors. I use small MOSFETs for the current source.
Note also that the collector-base capacitance of the driver is then bootstrapped by the signal. Obviously, don't try this at home if you are not familiar with HF stability issues in such a connection. This circuit also has some very big advantages under fault conditions.
I used this technique in the amplifiers used in my Athena active speaker systems, and a few more details of this amplifier design approach are touched on in the Athena ppt presentation on my web site at www.cordellaudio.com under "Loudspeakers".
Bob
darkfenriz said:
Adam,
That's right. The power MOSFET turn-on gate drive is provided by a current source that works against an emitter follower driver of the opposite sex to the power MOSFET. Thus, the driver acts to actively turn off the output transistor, rather than turn it on. If you play your cards right, the driver collectors can be terminated on the output line, so they see low collector voltage and can be fast devices that dissipate little heat. The current source that supplies it all need not be fast transistors. I use small MOSFETs for the current source.
Note also that the collector-base capacitance of the driver is then bootstrapped by the signal. Obviously, don't try this at home if you are not familiar with HF stability issues in such a connection. This circuit also has some very big advantages under fault conditions.
I used this technique in the amplifiers used in my Athena active speaker systems, and a few more details of this amplifier design approach are touched on in the Athena ppt presentation on my web site at www.cordellaudio.com under "Loudspeakers".
Bob
john curl said:
Good point, John. Hard to believe it's true, that they'd use such an old piece of garbage IC. Even a TL071 will beat the pants off of that IC. But then that player will also have horrid THD-20 as well. It's not a PIM issue, its a consumer garbage issue.
BTW, I've got a nice little Pioneer DV563A player that plays all sorts of formats as well. I've never looked inside. Maybe I should. However, I have noticed that it seems to spit out some low-level HF stuff in the SACD mode. Does yours do that?
Bob
Re: Re: Practical measurement of PIM
Bob,
thank you very much for the hint. And congratulations to your web page. I have the files already downloaded, now I have to get through the article in detail.
I agree, probably only a combination of methods can extend dynamic range of our measurements to the needed value.
Regards,
Pavel
Bob Cordell said:
Bob,
thank you very much for the hint. And congratulations to your web page. I have the files already downloaded, now I have to get through the article in detail.
I agree, probably only a combination of methods can extend dynamic range of our measurements to the needed value.
Regards,
Pavel