The noise is pretty high in the KH model. The -4402B used better opamps and might be a tad better. I found you can null the 2H with the recommended control... and a change from LM318 to LT1468. -125-ish. But the 3H can also be trimmed by shunt P pot (100K) acroos R186. No other harmonics are affected. The output stage is fine tuned for the 318 opamp... to cancel 2-3 Harmonic. So you cant change the output opamp.... but you can bypass it at the Inverted port on the front panel and use a high quality pot to reduce level down from 7 vlts. Thats about it for the 4400.
I can test any frequency within range of the analyzer and the AG16 source can be dialed to what ever you like. I'll post some numbers soon at higher and lower freqs per RC's request.
The approach that Demian is working on is really the most practical and cost effective means for DIY testing available today. IF you need much better than -100dB numbers. Commercial apps/designers would have to spend a lot more money.
Thx-RNMarsh
Last edited:
The 725-AG16 combo performance is pretty consistent to about 5KHz and the floor increases steadily from that frequency. One really neat feature is that the harmonic analysis works up to the highest frequency (100 KHz) and down converts the spectrum to 466 Hz (or whatever) so you can see through 1 MHz harmonics with a 10 KHz spectrum analyzer. I have to try one of the iPod app spectrum analyzers with it. It would be really convenient, and something to do with the obsolete iPod touch.
My experience with Japanese instruments, and products in general, if there is a clear performance spec, in this case distortion, the execution and performance is really excellent. Japanese receivers of the 70's and 80's all had much more accurate RIAA than American products for example. The 725-AG16 is an example of this single minded pursuit of clear goals
Oscillator mods -ranking:
yes. But in RF work it is alive and well.
The rank order for the oscillators with small but effective changes is from best to worst:
AG16 from ShibaSoku (stock).
HP339A (with IC mods and some Cap and pot. parts changes, plus null trimming mods)
KH-4402B (with IC mods)
KH-4400 (with IC mods and trim pot adj mods)
Stock, the KH4402B is twice or 6dB lower thd across the board compared to the earlier 4400. It has all the circuitry changed EXCEPT the oscillator IC. Its still a 318 opamp.
Just changing it out for a LT1468 which dropped the thd by another few db BUT the noise floor dropped by 10dB ! Now you can clearly see the 2H and 3H dominate and also higher order harmonics which had been covered by the IC noise.
I tried many of the opamps that are popular here in DIYAUDIO forums but only the LT1468 works best as a drop-in replacement. I included today the OPA604... its was worse than 318 in thd for this circuitry.
The AG16 comes in at worse-case frequency at -130dB. All other freqs are lower than -140dB. The clear and undisputed champion of super ultra-low distortion bad boy of audio signal generators is the ShibaSoku AG16.
That little HP339A is pretty darn good, however, when tuned up.
Thx-RNMarsh
yes. But in RF work it is alive and well.
The rank order for the oscillators with small but effective changes is from best to worst:
AG16 from ShibaSoku (stock).
HP339A (with IC mods and some Cap and pot. parts changes, plus null trimming mods)
KH-4402B (with IC mods)
KH-4400 (with IC mods and trim pot adj mods)
Stock, the KH4402B is twice or 6dB lower thd across the board compared to the earlier 4400. It has all the circuitry changed EXCEPT the oscillator IC. Its still a 318 opamp.
Just changing it out for a LT1468 which dropped the thd by another few db BUT the noise floor dropped by 10dB ! Now you can clearly see the 2H and 3H dominate and also higher order harmonics which had been covered by the IC noise.I tried many of the opamps that are popular here in DIYAUDIO forums but only the LT1468 works best as a drop-in replacement. I included today the OPA604... its was worse than 318 in thd for this circuitry.
The AG16 comes in at worse-case frequency at -130dB. All other freqs are lower than -140dB. The clear and undisputed champion of super ultra-low distortion bad boy of audio signal generators is the ShibaSoku AG16.
That little HP339A is pretty darn good, however, when tuned up.
Thx-RNMarsh
Last edited:
yes. But in RF work it is alive and well.
The rank order for the oscillators with small but effective changes is from best to worst:
AG16 from ShibaSoku (stock).
HP339A (with IC mods and some Cap and pot. parts changes, plus null trimming mods)
KH-4402B (with IC mods)
KH-4400 (with IC mods and trim pot adj mods)
Stock, the KH4402B is twice or 6dB lower thd across the board compared to the earlier 4400. It has all the circuitry changed EXCEPT the oscillator IC. Its still a 318 opamp.
I tried many of the opamps that are popular here in DIYAUDIO forums but only the LT1468 works best as a drop-in replacement. I included today the OPA604... its was worse than 318 in thd for this circuitry.
The AG16 comes in at worse-case frequency at -130dB. All other freqs are lower than -140dB. The clear and undisputed champion of super ultra-low distortion bad boy of audio signal generators is the ShibaSoku AG16.
That little HP339A is pretty darn good, however, when tuned up.
Thx-RNMarsh
Hi Richard,
Thanks again for this good information. If the spec for the AG16 is 0.0005%, that corresponds to -116 if I did my math right. So I'm guessing that their spec may be THD+N. Otherwise, on a spectral basis, they would appear to be doing much better than their spec.
What was the worst-case frequency for the AG16?
Cheers,
Bob
Hi Bob,
I think the spec is with a 600 ohm load and at a higher voltage than what Rick is using.
From what I gathered it is 5Vrms @ 600 ohm, 0.0005% THD.
Cheers,
Here is a measurement of the AG16 (My other, older version has the same performance, just no GPIB interface). The limit is the AK5394a in this case.
The spec is THD+noise. A quick trip to an online calculator shows the resistive noise floor of 300 Ohms (600 source and 600 termination) to be -128.367 dBV or .4 uV in a 30 KHz band. It becomes the limit for a broadband measurement.
I think this is characteristic Japanese instrument understatement.
The spec is THD+noise. A quick trip to an online calculator shows the resistive noise floor of 300 Ohms (600 source and 600 termination) to be -128.367 dBV or .4 uV in a 30 KHz band. It becomes the limit for a broadband measurement.
I think this is characteristic Japanese instrument understatement.
Attachments
OK I'm back. After leaving everything turned on all day.......
Into high Z load and 10 v rms and at 10KHz with 100Khz filter In -- no attenuation on output -- the ShibaSoku looks to be between .0002 and .0003% THD + N. 2H and 3H are down from those numbers (without the + Noise).... about -125-130 for 2H and another ten dB below that for 3H. [My guess is that I have about a 5dB total uncertainty. The instruments have not be freshly calibrated. AG16 might be better than I see if it was tuned up... The dominant is 2H... The others are after a tune up]. I am not using an ADC..... using the HP 3580A Spectrum Analyzer. It doesnt produce spurious harmonics at all on the 1v input range its being used. It is used on the distortion monitor output port of the analyzer.
PS -- THD is better at lower frequencies and the settling time at 20Hz is short....about 5 seconds by my count.
Thx-RNMarsh
Into high Z load and 10 v rms and at 10KHz with 100Khz filter In -- no attenuation on output -- the ShibaSoku looks to be between .0002 and .0003% THD + N. 2H and 3H are down from those numbers (without the + Noise).... about -125-130 for 2H and another ten dB below that for 3H. [My guess is that I have about a 5dB total uncertainty. The instruments have not be freshly calibrated. AG16 might be better than I see if it was tuned up... The dominant is 2H... The others are after a tune up]. I am not using an ADC..... using the HP 3580A Spectrum Analyzer. It doesnt produce spurious harmonics at all on the 1v input range its being used. It is used on the distortion monitor output port of the analyzer.
PS -- THD is better at lower frequencies and the settling time at 20Hz is short....about 5 seconds by my count.
Thx-RNMarsh
Last edited:
I think the EMU0202 uses a Crystal DAC and ADC. I have one and an 0404. The 0404 is better (it uses the AK5394a) but the implementation doesn't meet the same performance that the demo board does. Getting it all working was a chore as well, the drivers did not work as expected and it seems to be discontinued.
Thats about 4 dollars and 29 cents, I think. Anyway, it isnt as much as it looks.
The alternative is to use a few more of Viktor's oscillators. Maybe 4 to cover the range and put them in a switch selector box with one output. They are noisier but the harmonics are right down there with the Ag16.
THx-RNMarsh
Thats down below a new KH. If I can find one, I'll buy it and be done with this aspect of testing for awhile. Know who sells them -- in English?
-RNM
No. That's the only one that came up in Google.
You won't find many technical people in Japan who don't speak at least some English.
Just send an email.
http://orutika.com/e/detail.php?detail=011945
I'd pursue it but I like building my own.
Last edited: