BC550 BC560 Very low noise RIAA

Jump to Latest
#385
Er - no they don't. Farnell comes up with "No Longer Manufactured" and no method to order.

After placing a back order with RS, they have just sent a message "Thank you for your order, we apologise but we are unable to despatch the quantity ordered as this item has now been discontinued."

But thanks for the heads up on FindChips...

Not so good though. All the Digikey links come back with zero stock. And Arrow have a stock of a single BC560

Craig
 
Last edited:
#389
rsavas said:
LSK489 is typ 4pF, 8pF max Ciss, designed specifically for low input C
Click to expand...

You are absolutely right! Even Crss is very low at 3pF, and so impacts minimally even without cascoding. Add to that a 1/f doubling at an extremely low 10Hz and this looks like a really nice product. And it is available.

Also it is a monolithic dual, so inherently a matched pair with perfect thermal tracking.

There was a hint at one stage they were going to introduce a p-channel version, the LSJ689 (they released a datasheet), but nothing has appeared as something you can buy yet.
 
#390
sawyers said:
There was a hint at one stage they were going to introduce a p-channel version, the LSJ689 (they released a datasheet), but nothing has appeared as something you can buy yet.
Click to expand...

I bought one batch of 689s direct from Linear Systems in May of last year, and another batch from NAC in St. Pete, FL a month or so ago. I've been listening to them in an MM phono preamp for several months now. They are, I assure you, quite real.

My first batch of 689s ranged from 16 mA down to 8 mA IDss, where my 489s range from high 8's down to low 3's. I haven't characterized any of them for noise or Ciss.
 
#391
MarcelvdG said:
When you use series feedback, it is essentially only the gate-drain capacitance that loads the cartridge (1.8 pF typical for a J309 at 10 V VDG). With enough gain in the rest of the loop, you don't get any Miller effect worth mentioning either.
Click to expand...

When you say that when you have enough loop gain, it reduces the Miller capacitance of the input stage - do you have a reference? Thing is Marcel, I know from your publications and reputation that you wouldn't make a comment like that without some very serious thinking and evidence behind it.
 
#392
rco3 said:
I bought one batch of 689s direct from Linear Systems in May of last year, and another batch from NAC in St. Pete, FL a month or so ago. I've been listening to them in an MM phono preamp for several months now. They are, I assure you, quite real.

My first batch of 689s ranged from 16 mA down to 8 mA IDss, where my 489s range from high 8's down to low 3's. I haven't characterized any of them for noise or Ciss.
Click to expand...

Thanks for the heads up. They have not found their way across the pond yet. NAC have a modest stock in through hole and SM.
 
#394
Cis 5pF, Crs 5pf. Other specs look sensible for the application. Depending on the stage gain, Crs might be dominant. Marcel might have ways of mitigating that, although cascoding is the traditional way.

The LSJ289 looks good. Vn 2nv/rootHz, excellent low frequency noise, Cis 8pf and Crs 1.5pF at Vds 15V and Vgs 0V (so operating at Idss).
 
Last edited:
#395
And of course the monolithic duals LSK489/LSJ689 have excellent noise performance and at typical operating voltage low Ciss (few pF) and very low Crss (1-2pF). At those low values for Crss layout stray capacitance has to be carefully taken into account.
 
#398
One of the few manufacturers to quote rbb. The only thing that is missing from the somewhat scanty datasheet is 1/f.

Their spice models make sense from a noise perspective.

With zero input resistor, the model predicts 0.51nV/rootHz, which is about right for a base spreading resistance of 25 ohms. the 1/f doubling is sub 10Hz

With an input resistance of 10k (about what a MM is because of the big inductance) the PNP that model predicts:

10mA, 66nV/rootHz. 1/f 140Hz
1mA, 26nV, 112Hz
0.1mA, 15nV, 32Hz
0.07mA, 14.6nV, 27Hz
0.05mA, 14.3nV, 21Hz

So somewhere between 50uA and 100uA seems about right for the That PNP.

Of course that is an approximation, because a resistor of 10k is not a cartridge (typically 1k and 500mH) and there is an RIAA response to be taken into account, and on top possibly A-weighting.

But is the THAT 300 series suitable? Yes.
 
Last edited:
#399
Used THAT's model on LTSpice for a new simulation.

Results are quite good, but on the THD specs the slightly better version continues to be the one using KSC1845/KSA992.

Noise is practically the same on all.

The question is what to do with the offset.

THD is the smallest without the DC servo, but offset is 32mV. This won't certainly be value on an actual assembly, because of the almost impossibility to pair the N and P transistors, and even so keep DC low. Not even using the THAT chip or a dual FET will keep DC offset close to zero.

Of course the output cap will block that from the next stage.

The other option, which is using an NP large cap on the feedback increases THD. It goes from 0.000 to 0.00, which is still a good result but not the lowest you can get.

So what should we do?
 
#400
KSC1845/KSA992 - be careful using spice for noise, because the models do not have 1/f. The THAT models do.

To see if a modal has 1/f, check in the model for KF and AF parameters. AF might be 1, but if 1/f is correctly modelled KF should be ~10^-16.
 
You must log in or register to reply here.
Top Bottom