Discrete phono preamp designs?

[Paul Ebert]

I find myself in need of a phono preamp. I've looked around and it seems that I can find designs based on unobtainium components or designs based on opamps. Maybe I shouldn't be concerned about it, but I would like the first stage, at the least, to be discrete. Can anyone point me to designs that use readily available parts?

If not, anyone want to take a stab at designing one? Unfortunately, I'm not knowledgeable enough to do so on my own.

Whatever solution I settle on will need to work with average output moving coils (0.5 mv) and not be too expensive.

Thanks.

 

[cromodora]

Lp797

First stage at least, discrete...

LP797
LP797 supply
About LP797

P.S. maybe this thread should be moved to the 'Analogue' section.

 

[maxpou]

Hi,
very easy to built. Maxpou

 

[mako1138]

There's Norm Thagard's design in audioXpress, Jan-Feb 2001.

JFETs: 2SK389, 2SJ109
BJTs: LS352, LS312, MPS6523, MPS6521

I thought about building this design, but decided to go with an op-amp based one (ESP) for the sake of simplicity. I haven't done the research, but if the original parts aren't available, there's probably compatible ones with higher noise specs. Or with some redesign, it should be possible to adapt the concept for all-BJTs.

I too would like a discrete phono preamp, and I think have the knowledge to adapt/design one... But I have many unfinished projects as things stand.

 

[jackinnj]

Hi,
very easy to built. Maxpou

I simmed it and it ain't pretty.

 

[CBS240]

Hi

In the LP797 the hard to find duel J-fet appears to have both devices in parallel. Couldn't one just replace the pair with a single low noise J-fet and re-bias for that particular transistor?

 

[jackinnj]

Hi

In the LP797 the hard to find duel J-fet appears to have both devices in parallel. Couldn't one just replace the pair with a single low noise J-fet and re-bias for that particular transistor?

You can use a matched SK170 pair -- mount in close proximity -- you can epoxy them together or use a tie-rap. I believe that Linear Systems will sell small quantities of the LSK389. The pair of '170's knocks the noise down (it is already very low).

The high gain JFET's also are somewhat high capacitance...

I don't particularly like the AD797's in this application.

Here's the Maison Pacific simmed with Linear Systems model for the LSK170 -- and some changes to the capacitor values. note that the deviation in the lower chart is +- 1dB -- this is only a sim, I haven't breadboarded the device yet.

[IMGDEAD]http://www.tech-diy.com/RIAA/Maison.gif[/IMGDEAD]

 

[unclejed613]

the APT/holman preamp used a discrete phono pre, designed like a discrete opamp with an interesting twist..... the diff amp consisted of a JFET at the noninverting input, and a bipolar on the noninverting input. the 2 volt DC offset caused by this mismatch was blocked by a cap(the DC wouldn't have hurt anything, since it would be blocked by the rumble filter anyway). it was an excellent low noise RIAA pre.

i have a schematic of it, but i don't have it on this machine, so i'll upload it later.

 

[john curl]

Today, it would be best to stick with a fet diff pair. Fet noise is lower, today, and any advantage will be lost.

 

[scott wurcer]

John is right, that design was a cute way to save short money on parts but there's no magic in the asymetric input stage. TH has definately moved on to better thing$.

I found something funny trolling around this week, the microphone modding guys think the Japanese 2SK170's are bad compared to the LSK170's. The samples of LSK170's I evaluated met or exceeded every spec on the data sheet, but I still can't figure how either would stand out from the other in an appropriate circuit.

 

[john curl]

I think that they just 'lucked out'. I have VERY good low noise examples of 2n5564 devices from Linear Systems, but I have info from an informed source of relatively bad 2sk389 like devices from the same source. Sometimes, even Toshiba makes a bad batch, themselves, but it is rare. In 1990 or so, they tried to go to 'ion implant' and many low noise Toshiba devices got VERY noisy for awhile.

 

[analog_sa]

Re: Lp797

First stage at least, discrete...

Thanks for the links. Masterfully presented. Nearly spilled my coffee.

 

[jackinnj]

I think that they just 'lucked out'. I have VERY good low noise examples of 2n5564 devices from Linear Systems, but I have info from an informed source of relatively bad 2sk389 like devices from the same source. Sometimes, even Toshiba makes a bad batch, themselves, but it is rare. In 1990 or so, they tried to go to 'ion implant' and many low noise Toshiba devices got VERY noisy for awhile.

have you tried the LSK389?

 

[PigletsDad]

I have some discrete designs based on the low noise Hitachi (now Renesas) bipolars. In the UK, you can buy these from Farnell or Dalbani, and I guess other mainstream suppliers will have them too.

For MC inputs, these bipolars are attractive, as they have a very low 1/f noise corner frequency. Although the white noise of the best FETs, run with enough current, is pretty good, all the ones I have seen have much worse 1/f noise than a bipolar.

In a flat frequency response amp, 1/f noise doesn't matter much, as the ear is not sensitive to quiet low frequency sounds. But the RIAA curve adds 20dB of bass boost, and 1/f can matter.

One of my designs saw series production as part of the Maclaren 402 and 602 pre-amplifiers in the 1980s, made in NZ and exported widely. In the US, they were imported by Audioquest.

 

[jackinnj]

In a flat frequency response amp, 1/f noise doesn't matter much, as the ear is not sensitive to quiet low frequency sounds. But the RIAA curve adds 20dB of bass boost, and 1/f can matter.

The ITU has a paper on the noise specification for Audio -- digested here
http://en.wikipedia.org/wiki/Standard:ITU-R_468

isn't it surprising that the noise frequency that we are most sensitive to is that of a screaming wife? (6.3kHz)

and you can find more on the ITU's website:

http://www.itu.int

 

[Sigurd Ruschkow]

Erno Borbely measured this simple RIAA amp. You can read all about how it performed (not good at all, IMO):
http://www.borbelyaudio.com/pics/Borbely-jfet1-web.pdf




Sigurd

Hi,
very easy to built. Maxpou

 

[Sigurd Ruschkow]

PigletsDad,

what bipolars are you referring to?




Sigurd

I have some discrete designs based on the low noise Hitachi (now Renesas) bipolars. In the UK, you can buy these from Farnell or Dalbani, and I guess other mainstream suppliers will have them too.

For MC inputs, these bipolars are attractive, as they have a very low 1/f noise corner frequency. Although the white noise of the best FETs, run with enough current, is pretty good, all the ones I have seen have much worse 1/f noise than a bipolar.

In a flat frequency response amp, 1/f noise doesn't matter much, as the ear is not sensitive to quiet low frequency sounds. But the RIAA curve adds 20dB of bass boost, and 1/f can matter.

One of my designs saw series production as part of the Maclaren 402 and 602 pre-amplifiers in the 1980s, made in NZ and exported widely. In the US, they were imported by Audioquest.

 

[jackinnj]

I For MC inputs, these bipolars are attractive, as they have a very low 1/f noise corner frequency. Although the white noise of the best FETs, run with enough current, is pretty good, all the ones I have seen have much worse 1/f noise than a bipolar.

MAT-02 and MAT-03 from Analog Devices -- these are still available

 

[PigletsDad]

PigletsDad,

what bipolars are you referring to?




Sigurd

PNP 2SA1084 (or higher voltage version 2SA1085)
NPN 2SC2546 (or higher voltage version 2SC2547)

HFE linearity is very good - basically constant between 0.5mA and 15mA. If you run them with Ic of about 3mA white noise is equivalent to a 20Ohm resistor!

They are fast enough (120MHz) that there is no problem.

They also have very high HFE values - the E selection is in the 400 to 800 range. This brings the base current down, and hence reduces current noise, so a single input stage will run optimally over a reasonably wide range of source impedances.

 

[jackinnj]

PNP 2SA1084 (or higher voltage version 2SA1085)
NPN 2SC2546 (or higher voltage version 2SC2547)

looks like noise is 0.5 nV/rt Hz

Also looks like they are being discontinued by Renesas -- or at least "Not recommended for new design" -- you can get 2,500 of 2SA1084ETZ from DK in the US for $0.14 each.