"All American" RIAA Preamp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
wrenchone said:
If you want to sing the praises of another topology, please start another thread and do your singing there.

I'm surprised the usual pundits haven't been all over me...

Seems the pundits have respected your request, civil and respectful considering some of the other threads I've seen here.

On the subject of high-gm JFETs in current production, you may consider BF862 from NXP, search Scott Wurcer's postings here for more info. The SPICE model shows noise performance similar to 2SK170. An eBay seller in FL (westfloridacomponents) was selling them for about $10/50pcs last year; he still has BF861B with lower gm. They appear to be in current production, and Digikey carries them. They are SOT-23 only which may or may not be an issue for you. Also they are specified for AM radio RF preamps so examine the 1/f corner, BF861B datasheet shows about 2kHz, BF862 simulation model about 200Hz.

For MM levels the noise performance of the JFETs you have are probably good enough especially single-ended.

I think the availability of JFETs in TO92 types is endangered as time passes, but the SOT types should continue to be available.
 
Selecting surface mount devices is no fun, which is why I have avoided them for this project. Somebody no doubt makes a jig for that purpose, but I betcha it ain't cheap. Also, given the heap of detritus that is my lab, I could easily mislay a strip of 100 SOT-23 FETs and have them lost to the world for months.... Thanks for the suggestion, anyway.

The Japanese FETs I was keen on replacing are strange birds, with a high Gm at relatively low drain current, and a low cutoff voltage. The substitutes I've been considering all have higher IDSS and cutoff voltage, and need pretty high drain current to shine. I'll have to look at the devices you mentioned to see their characteristics. I hadn't considered NXP as a possible source.

I have a bag of J105s, which are chopper JFETs with a hilariously low on resistance. I've been wondering what their noise performance would be if properly biased at a high drain current. They would need a cascode connection to avoid high power dissipation. They might be an interesting choice for an MC preamp, but it's not an urgent matter at this point, as I don't own a low output MC cartridge.
 
On another note, the NXF FET sort of violates the "All American" premise, but I'll let it slide since you can get it at Digi-Key....;)

Sad to say, I just checked the Vishay web site, and they have obsoleted the J110, though they are still carrying the J310. The moral of the story is to buy a few hundred interesting devices when you have the chance...

Mouser is stil carrying the PN4391 series from both Fairchild and Vishay. I just bought 200 pieces each of Fairchild and Vishay PN4393, as it is a workhorse FET for me.
 
wrenchone said:

I wanted to try and build an good MM preamp using nothing but JFETs that I could easily order in the US.

...

On another note, the NXF FET sort of violates the "All American" premise, but I'll let it slide since you can get it at Digi-Key....

It's hard to say how long Digi-Key will be carrying the NXP BF series. They have lower Vp and higher gm/C than your selections. BF862 has appeared in Linear Technology application notes.

You probably know of Linear Systems and their 2nd sourcing of the Toshiba 2SK, made in USA and supposedly they take direct orders, your selections cost much less.

I doubt the Vishay and Fairchild parts are made in USA but you can buy them easily through distribution here.

I'm with you on the SMD handling problem, this is forcing me to go directly from simulation to PCB for one-offs instead of hand-wired proto boards, I'm tooling up for toner transfer process. I have clamped SOT-23s to adapter boards reliably enough for grading and then taped the graded devices down on overhead transparency sheets, this is certainly more painful than handling TO92s. There are SOT23-6 sockets costing about $50 made for programming uCs .
 
I should have defined my area of interest more precisely at the start, as I definitely wanted to exclude all 2SK170s and clones, and any other exotic Asian part numbers. This consideration came out of a urination match here (diy audio) a couple of years ago regarding suitable JFETs for audio applications, which galvanized me to find out how much juice could be squeezed from parts of domestic origin (meaning the spec was originally established in the US). At the time, I especially had the PN4391-4393 series and the J110 in mind, though I later discovered that a small signal N-channel MOSFET had promise for use in the preamp second stage.

The JFET market in general has tightened up distressingly since I started this project a couple of years ago.

Admittedly, the way multinational corporations operate, it's hard to tell these days where a given part is fabbed and assembled. Fairchild is a perfect example, as they have scooped up the small signal part lines of several companies (most particularly National). They (Fairchild) are based in the US, but I have no idea at all where they actually cook their JFETs. Vishay/Siliconix may be right down the highway from me in Santa Clara, but from what a Vishay expat at my work place tells me, the JFETs may be fabbed in Germany. BF types were also originally excluded from consideration (partially because I know nothing about them and have zero parts on hand), though if somebody else wanted to do a similar project using BF-series parts, I'd be interested in the results, especially given Nuvistor's mention of the BF861/862. Too bad those devices are SOT-23 - I pity the usual home hobbyist trying to select them, though admittedly, the spread on IDSS is narrower than the usual US chopper-type part numbers. It may be possiblle to drop in an unselected part and dial in the desired operating current using the supplemental current source I describe.

Vishay is steadily reducing their JFET product offering. They dropped the J110, though they still carry the PN4391-4393 and the J309-310. This may be due to ROHS-type pains as well as wars over fab capacity.

Fairchild is still carrying the J110, though I don't trust their abbreviated data sheets . I have a hundred or so of their J110s, though, and I intend to try some. They may be the only game in town soon for a lot of JFET parts. Their PN4393s were hilariously cheap compared to the Vishay variety, though I haven't tried to fire up any for comparison. I bought 200 of each from Mouser a couple of weeks ago with a comparison in mind, as I use PN4393s all over the place. I also got some PN4303s for general purpose low current use, as Vishay has dropped that part, too.

On Semiconductor has a very limited JFET offering. I haven't bought any of their stuff recently.

That's about it for US companies, except for the botique houses like Interfet and Linear Systems.

A few years ago, the list included Fairchild, Motorola, TI, Intersil, Vishay/Siliconix, National (they were heavily pushing their discrete JFET offering in the 80's).

If you exclude the 2N7000 (which may work in the second stage application - I haven't tried it), there are two big sources for small signal MOSFETS, especially if you want complements - Supertex (US) and Zetex (UK) (Zetex orignially got their Mosfet process from Supertex in the early 80's, when they were still Ferranti). I simulated my project using the Zetex parts, as I have the models, but I actually constructed my preamp using the corresponding Supertex part.

I hope this clarifies my stance somewhat. I've proved that you cna use faily pedestrian /JFET types and get acceptable results for a MM RIAA preamp. The end product sounds pretty nice, and once I get my preamp repackaged (new case, with a discrete JFET active crossover inside), I'll take it to work for some more formal measurements. Obviously, you can take what I've presented here and apply it to any part you like - I hope somebody does just that. I've thought of using a 2SK170 front stage with a MOSFET second stage, as that combination has a reasonably low parts count and acceptable distortion. I have in mind a new set of open-loop passively equalized preamps for the college station where I DJ, since we are moving our location in about a year and will need to reconstruct our master studio.
 
Re: Re: "All American" RIAA Preamp

nuvistor said:


Seems the pundits have respected your request, civil and respectful considering some of the other threads I've seen here.

On the subject of high-gm JFETs in current production, you may consider BF862 from NXP, search Scott Wurcer's postings here for more info. The SPICE model shows noise performance similar to 2SK170. An eBay seller in FL (westfloridacomponents) was selling them for about $10/50pcs last year; he still has BF861B with lower gm. They appear to be in current production, and Digikey carries them. They are SOT-23 only which may or may not be an issue for you. Also they are specified for AM radio RF preamps so examine the 1/f corner, BF861B datasheet shows about 2kHz, BF862 simulation model about 200Hz.

For MM levels the noise performance of the JFETs you have are probably good enough especially single-ended.

I think the availability of JFETs in TO92 types is endangered as time passes, but the SOT types should continue to be available.


I found not one but two working Quantech's a couple of weeks ago so I plan to finally nail down the low frequency noise on some of the currently available FET's. You can find or make little conversion boards for SOT23 to TO92. This doesn't help with the sorting. If your circuit forces a constant drain current the need to sort is lessened.

I got the idea of checking out the BF862's from the guys over at LT. It has a pretty unreal gm/C figure of merit. My initial use was for a condensor microphone where the 1/f is masked by the bias resistor noise. It needs to be checked for use in phono apps.

There is no free lunch, most of these high gm FETs suffer from excess gate current at high Vdg including the Linear Systems LSK170, BF862, 2SK170, and J310. This is impact ionization, a short channel effect.
 
Re: Re: Re: "All American" RIAA Preamp

scott wurcer said:

...most of these high gm FETs suffer from excess gate current at high Vdg

My first use of BF862/861B will be resistive loaded at Vdg=1.2V, in a 2-stage high gNFB active RIAA for MM and possibly MC so Vd will have little AC component except possibly on pops/ticks. Look forward to your information, my only 1/f test option is in application. SPICE noise looks OK but most SPICE noise models are not reliable, JFETs seem to match datasheet better than BJTs.
 
Ex-Moderator
Joined 2003
Talking of J110, the linearity is poor; considerably poorer than LSK170 (for example).

Scott: Would you care to expand on your noise comment?
 

Attachments

  • j110diy.gif
    j110diy.gif
    43.1 KB · Views: 756
EC8010 said:
Talking of J110, the linearity is poor; considerably poorer than LSK170 (for example).

Scott: Would you care to expand on your noise comment?

Initially I saw the BF862 in a couple of LT app notes boosting the performance of low noise photo-diode circuits. I thought they were really onto something, it had 1/3 the capacitance and the same sub-nV noise of some of the exotics. When you use it as an impedance transformer (like a condensor mike) the source impedance is really high at low frequencies. A typical high-end capsule might be 50pF || 1G Ohm. This means the source noise does not approach the FET noise floor until the kHz region and furthermore the roll-off is an RC one which is twice as steep as 1/f. So if the FET had a 1kHz corner it would get buried in the source noise. I need to measure the noise at low frequencies in a low impedance circuit. As of yet I have only confirmed that their process seems very clean and there is no popcorn or other gross noise anomolies.

BTW there is "clean" 1/f behavior that is based on first principles where each device exhibits the same limit, as opposed to someone being able to make the same device with no 1/f noise. This can happen with JFETs as well as MOSFETs (where it is better known).
 
The J110, if biased with sufficient current, has enough linearity for low level work. The thrust of this thread was never J110 vs. 2SK170, but to find a circuit to get the best performance possible out of FETs like the J110. I'm not tellng anyone not to use 2SK/LSK170 - its a free world, after all (more or less).
When I get my preamp repackaged, I'll follow up with measurements using the Audio Precision analyzer available to me at work. So far, it has tracked my high level simulation results pretty well. Low level results may be limited by the noise floor of our particular setup, always a possble problem for a piece of analog test gear tied to a computer....By the way. it would be instructive to show the curves of the LSK170 next to those of the J110 - a lot depends on signal level and bias point.
 
I'm just about finsihed a new layout of the "All American" board, changing the PNP bipolar transistors from 2SA965 to MPSW56 and substit\uting a new fast shunt regulator for the series regulator used on the original board. If anyone is curious about the new regulator, I'll post a link to the thread in the solid state forum - simulations looked very promising.. This will all go in a new case with a discrete active crossover so I can ditch my discrete crossovers and go biamp. In the process f stuffing the new board, I can develop an post a selection metod for the fets I'm using for the gain stages.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.