GPP - Great Phono Preamp

[Cauhtemoc]

Originally posted by scott wurcer
Try .00001Hz, I have. Had to prove that some wierd noise eventually went back to 1/f, it did and physics was safe again.

Oh good, we wouldn't want you changing the laws of physics without telling the rest of us.

[Sigurd Ruschkow]

Syn08 -
what kind of SNR (rel 0.5mV for MC and 5mV for MM) 20 Hz - 20kHz would those curves translate to with input short circuited?


Looks like a dual mono design, yes?


Sigurd

Originally posted by syn08
Moving Magnet, gain @1KHz is 44dB

AD797 + AD797 + LME49710
OPA637 in the power supply

[syn08]

Originally posted by Sigurd Ruschkow

what kind of SNR (rel 0.5mV for MC and 5mV for MM) 20 Hz - 20kHz would those curves translate to with input short circuited?

Looks like a dual mono design, yes?

I don't think you can get this from the straight from the previously posted results. Anyway, the measured S/N's are:

0.5mV MC unweighted 20Hz-20KHz 71dB
2.5mV MM unweighted 20Hz-20KHz 84dB

To go lower, more JFETs need to be paralleled in the input stage. Just for fun (I don't think there would be any subjective improvement) I tried 4x2SK170 and the above numbers went down to 75dB and 88dB. Six or more 2SK170 in parallel, stability at very high HF (over 100MHz) becomes an issue.

Distortions are not really important here, I haven't looked closer, but it's anyway under 0.005%v at 1V output.

Not sure what you mean by "dual mono" in this context. The power input takes 24VDC from a wall wart, splits to +/-12V using a TLE2426 virtual ground and then each rail is fed to a Jung super regulator, common for both channels. Crosstalk is measured under -100dB (still need to take a closer look at this).

I spent the weekend doing comparative noise measurements with three other solid state phono preamps, one is a $100 model, one is a $3000 model and one is a $1500 model, all new. I'll post the results later today.

The best combination so far is AD797+AD797+OPA627 and 2xOPA637 in the Jung super regulator.

[syn08]

Here's the promised comparison. I won't disclose the manufacturer's brands and models, but only the current prices (brand new). Enough to know they are all brand names, while B and C are highly regarded in Stereophile and other reviews. I don't have the schematics for any of the A, B and C so I can only speculate about their internal architecture.

A - "Low end", $100
B - "Mid Fi", $1600
C - "High End", $3000

All S/N numbers are 20Hz-20KHz unweighted, 2.5mV input for MM and 0.5mV input for MC.

Looking at the MM results, the differences between models are significant. My assumption is that the low performers (A, B) are using a regular opamp input stage, hence the poor noise performance.

Looking at the MC results, it seems like B and C are using the same input stage (so does GPP) and switches the gain downstream. Model A seems to switch/add a low(er) noise input (bipolar discrete? it certainly doesn't seem to be a low noise JFET design) stage, the S/N difference between MM and MC is very small.

But what seems to be really interesting is the very small gap between GPP, A and B MC S/N. Now, you tell me if this gap is worth $1500 difference between models A and B?

Experimentally, GPP may reach the Model C performance if upgraded to 4x2SK170 in the input stage. But 2SK170 are close to unobtanium, so other ways have to be developed. Two questions arise here:

- Is the performance gap between 2SK170 and any other solution worth anything but a nice spec sheet?
- If such performances are really worth of, how to proceed from here?

Next steps are a MM/MC phono stage based entirely on opamps (and see if there's any subjective difference), and if so, then exploring alternatives to the 2SK170 input stage.

[Sigurd Ruschkow]

71 dB SNR rel 0.5mV is OK, and excellent for a first design, but I would not settle for that when designing such a fairly complex RIAA amp as yours. Especially as I would design an MC-only RIAA amp for low resistance cartridges (<10 Ohm) with <= 0.5mV output voltage.

You might want to try low noise parallelled bipolars for the input stage. THAT has some low noise types, or you can try the 2SC2545,6,7E.

The 2SK170 are not hard to get at all. I do not understand that you find them close to unobtanium.....even the 2SJ74 are easy to find.

I would say that if you have a low resistance low output MC cartridge needing maybe 90dB of gain a low freq, more SNR can come in handy.

There are much more than SNR - how does it sound is the most important issue.
Overload issues are also important.
RIAA accuracy of course.



SIgurd

[scott wurcer]

Originally posted by syn08
- If such performances are really worth of, how to proceed from here?

Transformers?

[syn08]

Originally posted by scott wurcer

Transformers?

I hate them with a passion

[Bonsai]

Syn08, there might be another possibility here. since the JFET you are using is ultra-quiet, it pays to put as muchof the verall gain in the quiet front end stage as possible. So, I would experiment with increasing the JFET + first op-amp circuit block stage gain. Tradeoffs are required, but if its low noise you are after, this is one way.

[Sigurd Ruschkow]

The attached input stage will give an SNR of about 77 dB rel 0.5mV 20-20kHZ unweighted with RIAA curve using 8 pairs of matched SK170/SJ74BL,
and a very low resistance feedback network and a 5 Ohm pot.

The last few dBs are very hard to achieve. With or without transformers....



Sigurd

[syn08]

Originally posted by Sigurd Ruschkow
The attached input stage will give an SNR of about 77 dB rel 0.5mV 20-20kHZ unweighted with RIAA curve using 8 pairs of matched SK170/SJ74BL,
and a very low resistance feedback network and a 5 Ohm pot.

The last few dBs are very hard to achieve. With or without transformers....

Sigurd,

Paralelling more JFETs is the obvious solution in an attempt to gain some extra SNR dBs. Unfortunately I have found significant stability issues when using more than four 2SK170, unless you add some gate resistors which are then, of course, pooping the whole party.

Coincidentally, the Nelson Pass Xono (which I highly regard) AFAIK uses four cascoded 2SK170 in the input stage.