Transformerless JFet mic preamp

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

chrissugar

Well-known member
Joined
Jun 4, 2004
Messages
1,315
Location
EU
I don't know when was this page updated, but I can swear that last time when I visited this site, this preamp was not here [only the one with the Jensen transformer].
I supose it should be something excellent because it is designed by mister Borbely.
The kit is not expensive at all, 150Euro for normal PCB, 200Euro for teflon PCB [one channel].
probably this is the answer to our transformerless needs.
http://www.borbelyaudio.com/eb103213.asp
Kit price list:
http://www.borbelyaudio.com/kit_pricing.asp

PRR, would you care to analyse this circuit?

chrissugar
 
It's one of Erno's designs all right ;-) and that is not a bad thing.

Do you have a parts list to go with the schematic shown? I suspect the input FETs are the old favorite and still available (gott sei dank) Toshiba 2SK170/2SJ74 parts, which are wonderfully low noise, low pinchoff voltage. But what the cascoded FETs are I don't know---they have to have a pinchoff voltage magnitude larger than that of the input devices to produce enough drain-source voltage for them to operate, and have as well saturation drain current equal or better. I could rummage through some issues of Audio Amateur/Audio Electronics/audioXpress to find some Borbely articles and see what these parts probably are.

The attempt to do a fully complementary amp throughout is well-intentioned and largely successful, although Nature is cruel to us here, owing to the considerably lower mobility of holes compared to electrons in silicon. This means that, everything else equal, a P-channel FET like Q3,4 in the schematic will always have to be larger area and higher capacitance to have tranconductance and noise performance comparable to the N-channel devices. So there is an inherent asymmetry in the capacitances. Is this going to be a problem? Probably not---the input C's are doubly bootstrapped, the gate-source C's by the series feedback, and the drain-gate C's by the upper cascode FETs. And the voltage swing at the upper Q drains is to some extent reduced by the overall feedback as well.

Although the noise performance is exemplary (note how it degrades with even only a 50 ohm source Z---and that's not due to bias current shot noise folks, just 50 ohm thermal!) I believe there is a slightly more complicated topology that can reduce it even a bit more. Is it worth it? Well, probably not---but did the Audi A4 owners out there feel a burning desire to upgrade when the 3.0 liter engine models made them feel hopelessly emasculated in their mere 2.8 liter cars?

Erno clearly deserves his tariff on these designs, and helps to keep the community aware of the really great FETs. The semiconductor majors I am told gave up on the highest quality in these parts when they adopted ion implantation as the process of choice for impurity doping, in lieu of diffusion. It had little impact on most users and probably shaved a few hours off of the process, but the semi lattice and the resulting parts were never the same, despite the annealing cycle to try to put humpty dumpty back together again. The Toshiba parts continue to be fabricated with great care using diffusion I believe, but their worldwide comsumption is dwindling and parts are being discontinued left and right. To the possible rescue comes Linear Integrated Systems, who have just started to make a 2SK170 substitute and have plans for a dual part to replace the discontiuned Toshiba 2SK389.

Save the FET's!


Brad
 
Hey Brad

Thank you for your coments. I read the Borbely articles at least ten time each :grin: and there is a similarity with his other projects. He use 2SJ103 and 2SK246 for cascodes and for the low noise high Gm fets, as you say the 2Sj74 2SK170 pair.
I'm very interested in Fet opamps/circuits like the Borbely and Fred's http://www.forsselltech.com/JFET%20Opamp.PDF because I built some and I'm very happy with them.
I bought some hundreds of 2SJ74, 2SK170 and the duals 2SK389, 2SJ109 so the FETs are no problem for a while. :grin: Although I want to use them with care only in great applications.

chrissugar
 
Thanks Chris. Yes I have my hoard of FET gold too... It is a great relief to know that Linear Integrated Systems (from whom I receive no promotional consideration btw =] ) is starting to make comparable parts. I don't know that they will be persuaded to make P ch complements though.

I was fortunate to pick up some parts that were being discarded when Harman was shutting down a program years ago. They are the 2SK364's which although characterized for analog switching have the same chip inside as the 2SK170. At one point I had an assistant measure Idss and Vp and sort out matched pairs. There are some apps where the shared substrate of the dual parts is a disadvantage.

Then I bought a bunch of 2SK389 and 2SK109 just after they were phased out, so those will hold for a while. Got to remember to put those in the will---don't want them falling into the wrong hands.

Brad
 
where are you finding the 2sk170/2sk389 for cheap enough to buy hundreds?? I have found them at Newark(USA) but they want .89/3.09 for the damn things, plus a 10 day lead time!


:?:
 
I bought those hundreds at diferent periods of time in the last two or three years.
Most of them were bought from a local dealer. I bought every time all what they had.
Recently I bought from WWW.BDENT.COM [thank you Fred for the sugestion] 200 pieces of 2SJ74. I decided that every time I will have some spare money to spend, I will buy more Toshiba FETs.

chrissugar
 
> PRR, would you care to analyse this circuit?

No. I see FETs as tame devices, as they were 30 years ago. These particular devices are MUCH better than the ones I used to work with. At a glance I suspect something very good coud be done with them; and I'm sure Erno has.

I do agree that the new mike preamp kit look like a great tool. He has a couple variations on the theme sold for different applications.

FWIW: I just bought an Erno design. The 1980 Hafler DH-101 was designed by Borbely (and produced near my old stomping-grounds, and a few miles from where I bought it). It was well-regarded in its day. It is fully complementary, but all BJT (see? I was not the only one unthrilled with FET performance in those days). One thought is to add input iron and use it as a stereo live recording preamp. But I'll probably just team it with some obscure british gold-ear power amps I rescued and replace the $20 Japanese integrated monitor-amp that drives my KEFs. Afer all, if I like it, I can get another DH-101 for a song.
 
probably this is the answer to our transformerless needs

I think the answer is found already, and comes from a very nice DIY pal that always post here:

http://www.forsselltech.com/JMP-1.htm

Borbely make great amps, but I would go for this one from Fredd Forssell instead.
 
Yes Rafa, I totally agree, Fred is our friend and we should suport him. The JMP1 is a completed project, buy it, put it in a box, no headache.

Borbely has a schematic and we can talk about it, understand it, learn from it. This is what I mean when i say, it can be our answer to the transformerless pre.
We had a big thread about this subject but almost no real solutions. At least here we have a complete schematic.

chrissugar
 
Yes chriss. It´s missing the component values, but it´s something to start with. Maybe this thread will end with one more DIY group-schem, I hope. Would be nice to have a good quality JFET transformerless pre option as one of the GroupDIY projects. I just wanted to point to the Forssell preamp as a ready to go solution, because that´s what´s Borbely one linked in this thread seems to be. You know, just more options. The Forssell one is a little less cheap at $150 but I woudn´t say it´s expensive for what it is. There are three opeamps in there and the price is just a little more than the three separate opamps...

Just a question about the borning project... If we use many many fets in parallel, could we go with a cheaper one than the 2sk170, like the 2n5457 instead, and maintain good enought noise figure?
 
What is more troubling with good FETs is their general lack of availability. Due to the complexity of design required for high quality I am afraid that not many people would embark on building a fully discret transformerless FET preamp. Just mention measurement and matching of FETs and 98% of the builders on this board may shy away. Yes it is not as difficult as some think, but it is a major deterrent.
A hybrid approach like the Forssell version of the Cohen preamp is a really good alternative. I know that there are OPA2604 chips in the circuit, but when you match the input impedances for OPA604 types they sound outstanding. The dull character people associate with them comes from using them in non-inverting mode with the input impedances unmatched. This is a NO-NO with P-Channel opamp inputs. They are also excellent in inverting mode where the capacitance modulation is not an issue. Besides the monolithic dual opamp is perfect for low offset as the two opamps onboard tend to be matched naturally. Try to do that with discrete opamps. No wonder Fred charges $250 for that thingy he makes.

Just My Two Bitz,
Tamas
 
[quote author="rafafredd"]Just a question about the borning project... If we use many many fets in parallel, could we go with a cheaper one than the 2sk170, like the 2n5457 instead, and maintain good enought noise figure?[/quote]

Good point. The 2SK170 has a good complemetary pair the 2SJ74. The 2N5457 does not have a good counterpart.
If we take the theoretical input voltage noise of the 2SK170 at 1nV and the 2N5457 at 3nV (all at 1kHz) we will need about eight 2N5457s to match the Toshiba FET's performance. So instead of two transistors you need sixteen. Depending on your situation there could still be savings.

The worst part is that there are no easily obtainable FETs on the market that match the transconductance and Vgs of the 2SK170s.

Tamas
 
Yeah---the low Vgs is really an achievement (in fact it is so low that the nominal zero-tempco operating point is at a positive gate-source voltage (!)). Also I think you will find the low frequency noise corner much lower with the Toshiba parts than anybody's 5457's. Gate leakage, though not usually a critical audio app parameter except for condenser mics, is also extremely low in my experience for such a large-area device (also, the 170 would not be a very good match to any but a very high C capsule imo).

Like I say though, these guys in Fremont CA, Linear Integrated Systems, are serious about making at least the N parts, and have figured out how. I talked to them about the upcoming duals, nominally a -389 replacement, and the guy mentioned that they were an interdigitated construction which should be better for offset voltage and (he said) noise as well.

The 170 substitute is also available in surface-mount, FWIW.

Brad
 
Brad, what is the indicative of this 2SK170 substitute. Do they have plans to make a non SMD version?

chrissugar
 
The part number is LSK170 and it is also available in through-hole. They sent me a couple of samples right away, and they were responsive when I asked various questions. Here is the link to their site: http://www.linearsystems.com/

Brad
 
ok I love using FETs in all forms, especially since i use them daily in power controls. I'm trying to learn more about designing FET preamps in my spare time but have a few noob questions about them that I can't seem to find solid answers to.

In BJT preamps it seems to be the norm to have multiple stages, with the first stage usually buffering or at least with very low gain, then the next stage(s) are where the large increase in gain takes place. Would you want to do this with FET preamps also? I tend to think I wouldn't *HAVE* to but I'm also afraid i might have overlooked something. any thoughts?

In order to lessen gain in a FET stage would it be better to set gain by feedback in some form or to limit voltage through the part, therefor limiting the voltage gain?

I'm sure i'll think of other questions that I'm positive everyone else knows the answers to..

thanks! :thumb:
 
Typically FETs are lower transconductance than bipolars unless they are gigantic (like power DMOS). They also have a rather different transfer characteristic, nominallywith the drain current varying roughly proportionally as the square of the change in gate-source voltage. In ICs FETs good for analog amplification tend to take up a fair amount of area so usually not many are used beyond the input stage, as in so called bifet amps. There are CMOS analog parts but they are generally inferior, especially for noise.

But with their typically lower transconductance there is less need to "lessen gain" with FETs in general. If you need to, use local feedback (a resistor in the source for example).

There are some good discussions of gain structure in amps in general around and about. None specifically come to mind right now but a web search should be productive. I think Jensen may have a paper on the 990 amp which goes into some detail, and there are other discussions. But in any case for voltage-feedback amps it is often necessary to limit first-stage gain to avoid stability problems (oscillation in closed-loop) if one also wants decent voltage slew rate. With bipolars this is usually accomplished with resistors in the emitter(s). With JFETs they are usually not necessary. They are undesirable for noise regardless, so should be avoided if possible, and if the amp is used at high closed-loop gain it usually is possible.

This subject is vast---solid answers may be elusive for a while, depending on your starting point :).

Brad
 
Thanks Bcarso,

Like you said it is slightly misleading trying to read everything you can about it. everyone has a reason why their way is better than someone elses way and it can be hard to figure out which ones are better(or even correct..) without trial and error, which in my case is hard to do based on my work and home schedules, so i read as much as possible.

I am used to driving FETs to saturation as hard and fast as possible as my main design goal working with power electronics. working in transconductent regions is a little weird for me, but i am learning!

If you remember where or what discussions are best, please drop the links here so I can read them.

another question: what do you mean a "resistor in the source", do you mean resistence from the source rail to the source of the (J)FET? so really a resistence in series with the source leg? I'm unclear how this could be feedback so I am sure this is incorrect of me to think.

thanks!

oh and sorry Rafafredd, I didn't mean to sound rude if i did :green:
 
Back
Top