Its a post digital analyzer. Your 725D is also post digital with the brick wall filter. My 725 is older and doesn't have the brickwall filters.
The Panasonic (and Kenwood, Meguro, Shibasoku) are really targeted at production line service. They work in R&D but have the full programming capability to handle high volume production to high standards.
The HP, Boonton and Amber were targeted at a common spec that seems to be a military spec judging from the customers I have seen for many of them. The other big market in the US is Broadcast/FCC compliance. The stuff that came out of the car sound mfrs. (Delco etc.) has usually been Japanese or Danish not HP or AP. Probably comes down to test time on a busy production line vs. meeting mil procurement requirements and support (and knowing how to get the contracts). The mil requirements would not need to address digital audio since they are for communications test systems.
The Panasonic (and Kenwood, Meguro, Shibasoku) are really targeted at production line service. They work in R&D but have the full programming capability to handle high volume production to high standards.
The HP, Boonton and Amber were targeted at a common spec that seems to be a military spec judging from the customers I have seen for many of them. The other big market in the US is Broadcast/FCC compliance. The stuff that came out of the car sound mfrs. (Delco etc.) has usually been Japanese or Danish not HP or AP. Probably comes down to test time on a busy production line vs. meeting mil procurement requirements and support (and knowing how to get the contracts). The mil requirements would not need to address digital audio since they are for communications test systems.
The VP-7723A is better suited for production line testing and sample testing. It has a lot of features which make complete testing easier andhas memory (battery) of the settings programmed.
The neg is the digital display has the usual LSB flicker and affects the next digit some as well. Great as a random number generator but not too useful otherwise. This limits the accuracy of very low distortion numbers. However, the monitor output can be used for detailed analysis at under .001%.
The built-in oscillator has only two adjustments... one for a thd null (which was spot on) and a level adjust for calibration of the output level. The osc uses LDR for level control.
The THD rise as freq rises at about 6dB/decade.
FREQ THD (via AD725D)
100Hz .00006%
1KHz .00014%
10KHz .0003%
This is the first oscillator that I've seen with an all discrete circuit instead of an opamp. So upgrading it will not be done. The osc uses a metal can dual jFET I havent seen before -- uPA71A-M.
This units low THD and large number of features and extensive use of surface mount devices makes it hard to upgrade. But it would be good to do so as most of the smd are opamps of old technology --- 5534/TL07x, LM318... the usual suspects. However, the THD is rather low and therefore, with the difficulty to upgrade, I will let it be. Overall a pretty good unit and especially good for automated measurement/production.
Thx- RNMarsh
The neg is the digital display has the usual LSB flicker and affects the next digit some as well. Great as a random number generator but not too useful otherwise. This limits the accuracy of very low distortion numbers. However, the monitor output can be used for detailed analysis at under .001%.
The built-in oscillator has only two adjustments... one for a thd null (which was spot on) and a level adjust for calibration of the output level. The osc uses LDR for level control.
The THD rise as freq rises at about 6dB/decade.
FREQ THD (via AD725D)
100Hz .00006%
1KHz .00014%
10KHz .0003%
This is the first oscillator that I've seen with an all discrete circuit instead of an opamp. So upgrading it will not be done. The osc uses a metal can dual jFET I havent seen before -- uPA71A-M.
This units low THD and large number of features and extensive use of surface mount devices makes it hard to upgrade. But it would be good to do so as most of the smd are opamps of old technology --- 5534/TL07x, LM318... the usual suspects. However, the THD is rather low and therefore, with the difficulty to upgrade, I will let it be. Overall a pretty good unit and especially good for automated measurement/production.
Thx- RNMarsh
Last edited:
Twisted Pear is devopiing an ADC with the 32bit/384 kHz, -120 dB ESS 9102A A/D converter chip. It has the lowest THD I have ever seen speced.
http://www.diyaudio.com/forums/twisted-pear/186722-es9102-adc-module.html
If you could design an analog front end for that design, we would have a measurement system that beats everything out there today, and one that can get the data into the computer via USB.
Doing a front end with specs to fit the ESS 9102A is beyond my skills, but I will help out with whatever I can.
/S
http://www.diyaudio.com/forums/twisted-pear/186722-es9102-adc-module.html
If you could design an analog front end for that design, we would have a measurement system that beats everything out there today, and one that can get the data into the computer via USB.
Doing a front end with specs to fit the ESS 9102A is beyond my skills, but I will help out with whatever I can.
/S
The ESS 9102A datasheet´s performance figures are supposed to come form the eval board, so that means that :
THD+N @ -1 dBFS = -117 dB with 20-20 kHz BW (and Fs = 44k1) in stereo mode
THD+N @ -1 dBFS = -120 dB with 20-20 kHz BW (and Fs = 44k1) in mono mode
The plot in the DS showing THD+N vs freq @ -1 dBFS on an AP for a 1kHz tone shows all harmonics are <= -130 dB.
/S
THD+N @ -1 dBFS = -117 dB with 20-20 kHz BW (and Fs = 44k1) in stereo mode
THD+N @ -1 dBFS = -120 dB with 20-20 kHz BW (and Fs = 44k1) in mono mode
The plot in the DS showing THD+N vs freq @ -1 dBFS on an AP for a 1kHz tone shows all harmonics are <= -130 dB.
/S
I deal directly with a lot of chip companies. We have a variation on an old saying- when their lips are moving they are lying. . .
I want to see results from real parts. Looking at the specs they seem to be targeting consumer instead of professional applications.
Low noise doesn't mean low input Z for a voltage input, in fact it means high input Z. But I agree with thoughts that question the need for 32 bits.
Agree. If someone claims -120 dB THD+N in a ~20 kHz bandwidth I want to see a detailed measurement procedure how this figure is arrived at. As far as I know no currently available audio generator is specified to this level.
Samuel
R&S UPV-B1 claims -110 dB typ total for generator and receiver, so with
2 passive poles one should be about there.
Last edited:
Sory, but no as it is confidential. Do contact ESS and after signing an NDA, you will get the DS.
We pretty much agree. The equivalent impedance at the input would need to be very low. This can be determined by the source.
But I am not sure why you think low noise would mean a high input impedance.
Today A/D converters are getting to a real 24 bits at audio frequencies. However most of the 24 bit converters have 24 bits but the LSB's are not even close to accurate.
Now in doing an FFT analysis under some conditions it is possible to add uncorrelated noise (dither) to a signal under test and average it to get a small improvement in resolution. So under some conditions greater accuracy on a repetitive waveform may be obtained.
ES
The ShibaSoku AG15C is specified at -120 dB THD to 10 KHz but it doesn't say THD+N. Since they have a way to measure THD without "N" I guess that would be a way around this?
http://www.shibasoku.com/download/avc/ag15c_e.pdf
In sampling various osc and harmonic analyzers lately, none of them use the most current (Except newest A-P) opamps. Just changing out the old for the new opamps can improve the S/N and harmonic levels 20dB or more. So, if -120 is for an old design, it should be quite do-able to get to -140dB with best practices, parts and opamps. [notice I didnt say - easy - just do-able]
Thx-RNMarsh
Thx-RNMarsh
It may be possible but if some of the basic design decisions were made with the limitations of a 5534 then it may not be possible without extensive re-engineering. Reducing distortion products could be more possible through tweaking the AGC etc. But lowering the noise could require changes from end to end and still never get there. For example if the impedances around the tuning or AGC are too high the noise will never get there. Also the analyzer (necessary to see the results) may have higher internal impedances to facilitate protection from overload.
Also, its not clear that the 5534 really is a limitation on distortion when used optimally.
In short don't get your hopes too high. Something the an HP339 could go pretty far starting from a modest level or an ST1700. The newer units because of the automation and other design elements could be harder to get a 20 db improvement.
(I am mentioning this since I have spent way too much time pursuing these elusive goals with little to show for it.)
I think we need to start from scratch. Not only do we have control but we also gain a solid understand of the limitations. Tuning in an oscillator is limited to what the amplifiers can withstand for loading. This includes the total Z not just R. Strays can become a dominate factor in choosing. I suppose if we could build are oscillators with mini power amps much of this could be over come.