First I have to give one big thanks to people from diyAudio form who for giving me insight into the potential and possibilities for improving an already very solid pre-amplifier design.

Even as audio design is not my primary sphere of interest, the recent transition to LP records and my effort to do the de-claudization and return to analogue "value" starting form the beginning of this year, made me interested in the improvement of the existing gramophone pre-amplifier. the fact that I got a built-in preamp with the Pioneer PLX500 gave me the freedom to upgrade the ART DJ-PRE II that I get as a reserve.

After a thorough analysis of the original set, I noticed that it has a terribly prominent harmonic at 50Hz and its orders. I became interested in the improvement, so I attached it to my linear laboratory power supply and noticed that the harmonic can be suppressed, but also that the device can work with 6.5 - 7V DC, and that it gets heated on one side when connected to an AC power supply.

The test with the function generator showed a terrible tendency towards the appearance of harmonics of any dominant frequency (100,500,1kHz ...).

Keeping all the above in mind and knowing the tendency of fellow Audio "gurus" to create things "by ear" because in this engineering field 2+2 can be 6,613 etc. I decided to open the "thing" and see what it was doing. The first thing that caught my eye was the disproportionately poor design of the AC-DC conversion block, the Graetz bridge, then a small 470uF, then the Linear Regulator 7805 at 5V, and after that a small output filter, and only on the Op.Amp is it individually filtered, whatever arrives to them. An ordinary 7805 introduces no less than 50uV of voltage (AC phase) noise, which a good audio Op.Amp with SVR/PSR of -110/120dB easily rejects, but that is not the case here.

Why is it, because someone concluded that a simple CMOS op-amp with low en. of 4nV/Hz (at 100kHz!) is enough to get a solution that will be Low-Noise, "well, little tomorrow, maybe never". The ST TS972 has a typical SVR of -70dB which means it attenuates the noise level in the power supply by about 3162 times, before it appears as a parasitic value at the input of the Op.Amp. We all know that after that it will be 100 to 178 times more amplified. And the result of all that is the hum that can be heard and clearly seen on spectrographs.

The next characteristic of the TS972 is a THD (no noise, distortion only) of 0.003% which means that the first next harmonic will be 3.0E-5 weaker than the carrier signal which is only -90dB, if you think that is a lot add to this value and noise and gain and there you are in the audible range of below -70/80dB. But wait THD value is for AV=-1 only.

Since this Op.Amp needs to amplify the signal from the turntable whose resistance in the case of the MM head is below 1k Ohm, FET and CMOS solutions with low In in the fA/√Hz range are no advantage, moreover the high resistance at their input increases the thermal noise (Johnson-Nyquist noise).

So for a good solution we need any bipolar Op.Amp that can give at least SVR of -110/120dB, and below 8nV/√Hz at 10, 100Hz and even at 1kHz (and not in the ultrasonic range at 100kHz like TS972).

Also, the distortion should be displayed in the gain mode, not like the built-in ST TS972 where Total harmonic distortion is shown at f = 1 kHz, AV = -1!!!, RL = 10 kΩ is 0.003 % This THD is shown in ideal conditions for CMOS (like any other Op.Amp) since there is no Av=-1 gain and the output load is huge, nowhere near practical application for headphones or studio equipment set of 32,45...90 or 600Ohm (or even for 1k and 2.2 k Ohm according to RIAA spec.) This practically means that it is clear why the distorted images are such an expression because 0.003% at 1k becomes 0.03% and when it is amplified 20 times Av=20 ie. 26dB increases to 0.6%, i.e. -44.4dB difference between the carrier tone and its first harmonic.

So I took it upon myself to search the net to see if I was the only one who noticed these anomalies because the searches are full of praise and wonderful reviews full of praise for the -80dB characteristics of this pre-amplifier with also "little" distortion. So the device is even included in reviews and in the analyzes where it is compared with 100-200 USD more expensive models that show similar (in my opinion, frankly poor) characteristics.

All in all, since this post was long enough to make it a little shorter, I tore out all the Op.Amps, Ordinary electrolytic capacitors and thoroughly reworked the power supply with the necessary addition of components.

Since my idea was not to turn a 50 USD device into a 500 USD beast, I gave up on the OPA1612 or OPA2228 which were my primary choice, since the existing state of my inventory with 3-4 pieces I can't possibly replace due to the lack of components on the market and price hikes, I decide to save them for some other/new projects. Next I tested the LM4562 and its successor family LME497xxx and since I couldn't find a reasonably priced SOIC version I abandoned them also, I decided on the suggestion made by NwAvGuy in his analysis. Honestly my favorite for a cheap solution was the NE5532A at first, but it The NJM4580D outperformed it in my tests, and only one small thing decided that I should include the NJM2068D (MD) in the design instead, and that is the equivalent noise level, which in my tests is just behind the leading trio of OPA1612/1602, OPA 2228, and LM4562... even more form 1-10Hz up to about 400Hz its noise level is lower than LM4562 or OP227. LT1028 hash lowest values but i is best that is it hard to teme.

And it has specified and guaranteed RIAA caracteristics NJM2068D Equivalent Input Noise Voltage is as FLAT+JIS A, on RS=300Ω at 0.44 to max 0.56μVp-p

Spec do not show its en(Equivalent Input Noise) value but it si perform better then 3-4nV/√Hz devices on 10, 100 and 1kHz so it have to be below 4nV/√Hz.

Even its mediocre THD-N is shown at near real situation as sum of THD with Noise at AV=20dB gain amplification, VO=5V, RL=2kΩ, f=1kHz and it is THD-N is 0,001%

Simular new design NJM8068 (repackaging) spefied also as 0.001% even show that it can go as low to 0,0007% at AV=100 at 8-9V output.

When Ti put in spec for OPA1612 as stellar level for THD-N like 0.000015%, we must have in mind that it is mesured on no gain platform AV=+1 so that real condition Pre.Amp wil degrade that for any aplification level it performs, so at AV=100 (20dB relative) it is =

20x Log(THD-N/100)+ 20x Log(AV) ---> -136.478+40 => -96.478dB which is translated to NJM specs equal to 100% * 10^(-96.478dB/20) = 0.0015%

Now it is clear why on practical tests "King of the hill" is not always one that we all expect to be.

And the most important thing is when all supplier have stuff an waiting until next year, I managed to buy them for 1 USD each.

My mod list is the following:

Replaced blue LED with green LED which is much dimer

Added On/Off switch on chassis

Added power supply switch (to do this, I interrupted -cut- traces on the bottom of the PCB).

Added 2200uF/25V LowESR high ripple electrolytic Cap. (brand name)

Boost working condition by replacing 5V (Korean) for ST L7810A low noise 10V linear regulator

Added Lin. Reg protection diode for hi cap. load switching

Added another 2200u/25V LowESR high ripple electrolytic Cap. (brand name)

Replace all onboard ordinary electrolytic Cap. and replace them with audio grade (brand name) equivalent LowESR high ripple electrolytic Cap From Panasonic and Samwa.

Replace 1uF electrolytic on WIMA MKP in audio path and with MKT on power path.

Replace condensers for 100pF input filtering

Replace addition 100pF condensers for 200pF filtering with 220pF NP0 ones, so that now I have 0pF and 220pF input filtering

And most importantly I replace TS972 with little larger NJM2068MD (D-sufix version for guaranteed RIAA req.)

This post is becoming even longer so I will put all other stuff to my blog

All of that work will be explain in 4 modification phases (or maybe more) in this Project.

Best regards ,

dr Jovan I.