This is not true in many cases and as I have said before, your blanket statements are incorrect and not helpful. You seem to have an unhealthy attachment to LME series op-amps.
That's a weak reply, but I did anyway.
Self shows the AD797 as similar or lower distortion, especially with low impedance loads ~500 ohms, in his "Crossover" book- p.398 versus p.407.
Max input noise of the 797 is 12 dB lower (!).
Monty M. in this forum reports better results with the 797.
However, off topic now that Jens has explained his position- so if you have some actual evidence then please email me.
Best wishes
David
Last edited:
Self: "This is a remarkably quiet device in terms of voltage noise, but current noise is correspondingly high due to the high currents in the input devices"
I see THD 0.0007% at 20 kHz for LME, 0.001% for AD. Load is 500 ohms.
Last edited:
Self doesn't mention it but the current noise in the 4562 is not much better than the 797 - 1.6 versus 2.0 pA/rt(Hz) typical @ 1kHz (there's no max number on either spec sheet).
Whereas the spec sheets show max 4562 V noise is 12 dB worse than the 797.
So the 4562 has the down side of input currents comparable to the 797 but not much of the benefit.
I don't know how you can see this because it's not in the book, certainly not my copy at least.
LME4562 number is correct but p.407 only has one plot (13.25) and it shows 0.0005% for the AD797 at 20 kHz, load 500 ohms (and lower than the LME elsewhere too).
That's from the first print run, Focal Press, 2011.
In any case I won't respond any more here, post, if you must, in the "Low distortion oscillator" thread where you have also raised the issue.
@Jens- sorry for the off topic.
Best wishes
David
Last edited:
my comment re the RTX is that I use it for site visits and such, ie portable. The fact of it being integrated with virtens is very useful to me.
Kudo's on the proect and that should give you feedback about size and ease of use.
I don't always use it to the precision it allows (not needed), but knowing that is not limiting is wonderful. I do a lot of tape setups and troubleshooting for the music industry and it is very quick to use and powerful. (freq response, boas, wow and flutter, etc)
Cheers
Alan
Kudo's on the proect and that should give you feedback about size and ease of use.
I don't always use it to the precision it allows (not needed), but knowing that is not limiting is wonderful. I do a lot of tape setups and troubleshooting for the music industry and it is very quick to use and powerful. (freq response, boas, wow and flutter, etc)
Cheers
Alan
Very interesting results are here! I see THD -135 to -137dB at the -15 to -17dB level.
What is the ADC on this measuring? The RTX?
Did you vet the LME49720 before you tried it? I'm sure you know this, but around 30-40% of a batch will have issues with LF noise. It tends to be below -100, but is very obvious on an FFT. I haven't had a chance to see how bad it is in VF mode (I suspect it will be virtually impossible to see, but could be wrong), but it definitely manifests itself when gain is applied in Inverting / Non applications. I have tubes of rogue 4562s that exhibit the issue. It's typically below 100 Hz and slightly oscillatory in nature. It's been suggested that the issue is due to the fab process. The peculiar thing is that Self didn't mention this in his seminal book...nor have other esteemed designers I've spoken with mentioned it...
No, I didn't check the LME49720 before the test. And it seems to work OK. Only the OPA1612 showed a weird behavior.
Actually I haven't seen any problems with the LME49720's used in the Audio Analyzer. And I personally tested all RTX6001's made.
I assume that the post-DAC filter/buffer is not the most critical application, so maybe that is why I haven't noticed any problems.
I did replace other semiconductor parts during the RTX6001 production test because they were too noisy. But none of the LME49720's.
Actually I haven't seen any problems with the LME49720's used in the Audio Analyzer. And I personally tested all RTX6001's made.
I assume that the post-DAC filter/buffer is not the most critical application, so maybe that is why I haven't noticed any problems.
I did replace other semiconductor parts during the RTX6001 production test because they were too noisy. But none of the LME49720's.
I evaluated a couple of industrial grade 20bit SAR ADCs (LTC2378-20, ADS8900B) and definitely the audio grade ADCs are not worth fighting with. According to my testing, the LTC2378-20 from Analog Devices/Linear Technology gets the first prize with:
SNR 102dB
THD -128dB
SFDR 130dB
ENOB 16.45bit
Noise Floor -149dB
All up to and very close to FS (0.03dBFS) which is 10Vpp, BCLK=24.576MHz, Fs=BCLK/80=307.2KHz. Signal fo=2KHz, bin width below is 2.344Hz, about 20 averages.
The TI ADS8900B is close, but slightly worse (THD is -124dB, with the same OPA1602 op amps). I am myself also put off a little by the TI part datasheet, it is not as straightforward to use as the AD/LT part (although the multi SDO mode is interesting). For both, interfacing with I2S is, if not trivial, rather easy, at least for the LTC2378-20 for which I looked closely. A dual channel implementation (the part is single channel) appears also very easy.
SNR 102dB
THD -128dB
SFDR 130dB
ENOB 16.45bit
Noise Floor -149dB
All up to and very close to FS (0.03dBFS) which is 10Vpp, BCLK=24.576MHz, Fs=BCLK/80=307.2KHz. Signal fo=2KHz, bin width below is 2.344Hz, about 20 averages.
The TI ADS8900B is close, but slightly worse (THD is -124dB, with the same OPA1602 op amps). I am myself also put off a little by the TI part datasheet, it is not as straightforward to use as the AD/LT part (although the multi SDO mode is interesting). For both, interfacing with I2S is, if not trivial, rather easy, at least for the LTC2378-20 for which I looked closely. A dual channel implementation (the part is single channel) appears also very easy.
Attachments
Last edited:
PMA also noticed issues with this op amp, reported in this thread
Sensitivity of opamps to air coupled EM fields, especially of the LM4562/LME497X0 family | Audio Science Review (ASR) Forum
I don't know if this is related to your issue, or if his explanation is correct but looks like some sort of problem.
The ENOB is not what it could be.
Any comments?
Best wishes
David
That's what the measurement is telling me. Theoretically,
ENOB=(SINAD-1.76)/6.02
SINAD=(Psignal+Pnoise+Pdist)/(Pnoise+Pdist)=100.7dB (reported as measured)
So ENOB should come as 17.02 bit. I can live with the 0.57bit difference, though. My guess is the tool has a slight error in evaluating the SINAD, it is reported as exactly the same value as the S/N which is incorrect, S/N=Psignal/Pnoise.
Last edited:
The LME family was a victim of the TI takeover. I understand that TI decided that the premium audio stuff would stay with Burr Brown and planned to kill off the LME stuff (and dume most of National's talent) The LME was on a specific highly tuned process and was selected for leakage (the primary difference in the different dual opamps that started with the LM4962).
The current opamps are made at a different fab on a different process (similar I'm sure) but are not the same part under the plastic, much like a Signetics 5534 (the original) is not the same as a current JRC5534 or Phillips 5534.
I would not be at all surprised if there are significant differences and part variations. Most interesting would be to set up a jig for testing leakage at higher voltage (not really sure how to do that) to screen the parts. For the near term for my efforts I'll use my Quantech. Most of my parts are older from early production but I'll get some current ones once I have the test setup.
The current opamps are made at a different fab on a different process (similar I'm sure) but are not the same part under the plastic, much like a Signetics 5534 (the original) is not the same as a current JRC5534 or Phillips 5534.
I would not be at all surprised if there are significant differences and part variations. Most interesting would be to set up a jig for testing leakage at higher voltage (not really sure how to do that) to screen the parts. For the near term for my efforts I'll use my Quantech. Most of my parts are older from early production but I'll get some current ones once I have the test setup.
Lots of amazing op amps in this power supply range (10V, with 11-12V absolute maximum rating) check out ADA4899 with -117dBc @1MHz and 1nV/rtHz. I had some OPA1602 at hand and put them at good work.
Should be very close because the THD is so low that the SNR will dominate.
I am just a bit surprised SNR is only ~17 bits in a 20 bit chip.
I do understand that an industrial SAR chip won't match an audio chip optimised for SNR but haven't studied SAR sufficiently to appreciate the details.
Best wishes
David
I think you are confusing the DAC ENOB with the DAC Effective Resolution.
The Effective Resolution is defined as
ER=Noise free code resolution+2.7bit
where
NFCR=log2(2^N/Vpp_input_referred_noise_counts)
Therefore, when calculating SINAD and ENOB, the noise and distortion include not only the input referred noise of the ADC, but also the quantization noise and the distortion. Hence, the SINAD and ENOB are a measure of the dynamic performance. Effective Resolution is used to measure the noise of the ADC under DC input conditions; in this case, quantization noise is absent, so ER is always larger than ENOB.
In this case, from measurements, I get Vpp-noise (input shorted) about about 14 ADC counts (which is about 70uV given the 5V FS), therefore NFCR is about 16.2 and ER=18.9bit, much closer that what you would expect, correct?
Yep, we are using ADA4899 to drive SAR ADCs in a few projects at work. Another new part that’s interesting if you need the speed is LTC6228.
But high-ish 1/f corner.
