Help requested from JFET-knowlegable people :-))

[andyr]

I started to build a battery-powered, JFET-based phono stage mid last year. The circuit was sent to me by a friend of mine who said it really sounded great ... so I thought I may as well try it out, to see whether it bettered my own phono stage. So I built it and listened to it ... yes, it sounded very nice but, when I measured its RIAA equalisation (using a sig-gen fed through a Hagerman reverse RIAA board), the 20hz reading was low by over 4dB.

This phono stage uses:
* a pair of JFETs as the input stage, followed by
* a passive composite RIAA network, then
* another JFET gain stage, followed by
* a transistor output buffer (unity gain).

Schematic attached.

I found that the values of the components in the passive composite RIAA equalisation network do not match the values obtained from using the equations in this most interesting article:
Phono Stage Design

So I wanted to correct the values. I have subsequently twice rebuilt the RIAA equalisation circuit so that it matches the equations (each time making some different assumptions) and still can't get it to measure any better.

So I thought I would ask here for help from anyone who knows their JFET theory.

The first question I would like an answer to is this:

* the attached article shows that R1' (which is the value obtained by R0 (240K in the schematic) & R1 (510K) in parallel) is the basis for generating all the passive RIAA network component values.

* however, I was told that you have to add the Zout of the input JFET stage to R0, to get the right calculation for R1'.

Is this correct?

My second question is:

* is it true that the Drain resistor of the input JFET pair defines their Zout?

I appreciate that in the circuit shown, the Drain resistor value is equal to the source resistor value (so it doesn't matter) but, in my circuit the Source resistor is much lower (because I've used a DC rail using 5x6v SLAs = 30v, not 3x9v transistor batteries = 27v).

All I know is, when I calulated R1' by taking 291K & 510K in parallel and then generated all the RIAA values using the equations in the attached paper, it didn't measure any better than the original circuit.

So I'm wondering if the Zout of the input JFET pair is in fact equal to something else than 51K?

Thanks,

Andy

[wakibaki]

51k || impedance looking into the drain (g[oss]). This is typically in the range 100k to 1M (Horovitz & Hill) . So you might be looking at as low as 33k

w

[davidsrsb]

This data sheet:
http://www.onsemi.com/pub_link/Collateral/MPF102-D.PDF
shows the reverse transadmittance is about 0.01 mmhos ie 100k
Two in parallel is 50K, so the RIAA network is loaded down at low frequencies
This device is intended for rf work and is badly controlled for audio applications like this eg the decade range of Idss

[mjf]

hello.
if i read correctly there is a 0,01uf poly cap between the drains of the first stage and the 240k res........this cap is a little bit small, i think it cuts off the bass.
have you tried out to make it a little bigger........may say 47nf or 100nf?
greets

[revintage]

As mjf says the major problem is the coupling cap that is both wrong value and also wrong placed.

This is an example of a +/-0,1dB correction with the 4th time-constant added. Unfortunately there is to low current through the FETs. Adjust the source-resistor to get Ud just above 10V.

[AuroraB]

As said - I've seen the MPF102 used as RF amplifier several times, but never on audio. 2SK170 would be a much better choice....

Also - the inverse RIAA ciruit being fully passive, expects a 47k load -not 51k. This will affect the readings somewhat (without doing the calculations). It also assumes your test generator have an output impedance of 50 ohms. Not all of them have, and some are switchable. How about yours?

[revintage]

Hey AuroraB,

47k or 51k doesn´t matter that much even if it will give some small deviations. The correction together with the malplaced and to small couplingcap is the major source of trouble.

The one I show above is fed via a corrected version of Hagermans theoretical rev-RIAA. Unfortunately he made a mistake in his article which he admitted when I contacted him. His commercial revRIAA is correct though. But as you say it must be fed with the correct source and loaded right.

The FETs run at about 150uA each and that isn´t optimal in my eyes. SK170 at 5-8mA would be a far better choice.

[andyr]

Originally Posted by davidsrsb

This data sheet:
http://www.onsemi.com/pub_link/Collateral/MPF102-D.PDF
shows the reverse transadmittance is about 0.01 mmhos ie 100k
Two in parallel is 50K, so the RIAA network is loaded down at low frequencies
This device is intended for rf work and is badly controlled for audio applications like this eg the decade range of Idss

Hi David,

Sorry, I don't have too much knowledge about this ... are you saying that the MPF102 is the wrong JFET to use and I should replace it with one that's designed for audio frequencies?

Like the SK170 which revintage suggests or the 2SK170 which AuroraB suggests, in threads further down?

Thanks,

Andy

[andyr]

Originally Posted by mjf

hello.
if i read correctly there is a 0,01uf poly cap between the drains of the first stage and the 240k res........this cap is a little bit small, i think it cuts off the bass.
have you tried out to make it a little bigger........may say 47nf or 100nf?
greets

Thanks, mjf.

Yes, I had been told before that the input cap to the RIAA network was a bit low ... so I doubled it. But it seems I should've "ten-times'd" it!

But, given that the phono stage output coupling cap is 1uF, can you explain why 100nF will be sufficient for this cap after the input pair of JFETs - ie. if 100nF is good, would 470nF be sonically better?

Thanks,

Andrew

[Sheldon]

You didn't specify the signal generator used. If you are using a sound card, the output impedance may be too high. I had to buffer mine to get it to work properly with the reverse RIAA network

Sheldon