Solidstate Preamp Project Issues

Hey all, I have slowly been designing a solidstate preamp for over a year now in my spare time (not that much spare time..). I'm finally done, but there are a few nagging issues I have that I was wondering if folks could provide some help with. The basics: it is a dual JFET input class A preamp w/ 32V a bipolar supply. The output stage is a beefy MOSFET stage that has enough standing current to drive less than a couple hundred ohms at full output. The headroom is about 31dBu. I can get about 60dB of gain out of it. I have a working DC servo circuit, but have pretty low DC offset normally (of course it drifts a little). Also I use a Jensen output transformer for the balanced output. This was my first DIY pro audio project (even though I am an electrical engineer by profession) and I made assumptions in the beginning to make the design easier for me. So enough babbling, on to the issues. I will post a schematic at some point probably. (sorry for the long post, but i think i might raise some good topics here i split my issues up to different volumes)

1. Layout. I first breadboarded the circuit. I noticed while testing the unit with an AP analyzer, that referencing the breadboard to a ground plane reduced the noise by over 6dB. This was simply taking a piece of aluminium foil, placing it under a piece of paper which was under the board, and clipping the foil to star ground. I knew that there would be better electrical shielding once the board was in a metal enclosure, but I wanted to experiment with a ground plane on the board. I designed a layout and got a board fabricated with a ground plane. My plan was to compare the noise performance with a board with the ground plane to a board with the ground plane connections drilled out and individually wired back to a star point. I found that maybe the best solution is to individually route back ground connections to a star point, and have a ground plane connected only to star ground so that no part of the circuit uses the plane as a return path. I noticed many pro audio circuit boards have no plane at all. I have also read that some people do not use planes due to stray capacitances it can induce. The main thing I discovered is to definitely route back bypass capacitor ground returns to star seperately from everything else. Also a bigger cap here is not neccessarily better, as the charging current can cause hum in the rest of the circuitry. Thoughts on this? Anyone have anything to add? Also, what about audio transformer shielding, I haven't even looked into this at all? My noise performance is not bad. I think it was -120dBu referred to the input. The 60/120hz contribution was negligible. Of course I would like to improve the performance if possible. The majority of the noise came from the shunt feedback resistance. This leads me on to Volume 2: Gain Control. I'll post another topic soon.

And if you need web space to host the image, feel free to send them my way and I can host for you:

consul [at] studioconsul [dot] net

You just need to deobfuscate the addy. :green:

Hi soultek,

Welcome to "The Lab"!

Sounds like an interesting project- nothing like designing from the ground-up :wink: . The ground-plane issue is very interesting. I've been building my breadboard pre-amps on a piece of copper clad "dead bug" style and using the copper sheet as the ground plane with "islands" glued on. The whole setup is plonked into a biscuit tin which has one ground connection to the star-ground. This does help the noise floor a lot, and external interference is negligible!

I know that SSLtech (Keith) did a homebrew ground plane on his SSL 9k mic pre- you etch one double sided copper clad, and then drill through the component lead holes and then countersink the holes on the ground-plane side.

I've read quite a few Nat Sem app notes about decoupling caps- there's an article called "mixed signal techniques" or something like that which explains about ground-modulation from the charging currents to these capacitors in a circuit. The magic is in the details here I'm sure.

Anyway, looking forward to hearing more thoughts into the development of your design.

:thumb:

Mark

More issues I ran into with my project..

2. Gain Control. The method of gain control was a difficult balance for me, I wanted to be able to easily source quality parts, and at the same time get a circuit that results in high bandwidth, low noise, stable gain control.

I followed much of the advice in the Jensen ap note: "Tips on stabilizing opamps" which I found very helpful. I also looked quite a bit at the writeup on the JE990 too. My opamp would be non-inverting with the standard series/shunt feedback from the ouput. I decided to use a fixed series R while varying the shunt R. I also decided to use a cap across the series R to get some phase lag and buy some phase margin. I noticed that many designs place a very large cap in series with the shunt R which prevents any DC gain. To get a very low frequency response, you need to use a cap at least 1000 or 2000 uF. For some reason this bothers me, and I will get to that later. I planned to use a rotary switch, but finding one that I was satisfied with was challenging. Plus I wanted a non-shorting one so that between stops, the gain can default to a low setting. A shorting one would make the gain jump between stops, not good. I had a hard time finding a 12 stop, single pole, quality, reasonably priced rotary switch and I gave up. I did buy a couple from a local surplus electronics shop for prototyping. I then decided to use a quality conductive plastic pot. After discovering that it's amazingly hard to find reverse log pots for sale anywhere--even though people claim to make them still--I gave up and went back to a rotary switch just to build up something that works. My rotary switch switches in a resistor to parallel with the lowestest gain shunt resistance so that between stops it defaults to this. It works good functionally, but I still need a reliable source for these rotary switches. Help? Also the shunt resistance value varies from 10 ohms to around 600 (I have about 44 dB of gain on my first build of the opamp/preamp, I plan to use it as a line amp). At low gain settings, the majority of noise comes from these few hundred ohms of shunt resistance. I think the contribution is about 3nV/Hz^.5 Any clever ways to reduce this? I thought about using smaller shunt resistors, but it's hard to find 1% thruhole resistors less than 10ohms. and usually they are the 1W current sense types too. Of course i can parallel a bunch of stuff too. If I can drop down the Rseries/Rshunt resistances (and Cseries capacitance) by a factor of 10, I can make this thing damn quiet. I think it will be tough to get the stability and transient response right though. I'll forward my schematic to Counsul so he can post it for me. Comments on gain control anyone?? I'll try attaching the schematic here:

The schematic is available here:

http://studioconsul.net/techtalk/soultek/di.pdf

I haven't had a chance to take a close look yet...

I don't understand V3(1.6v), can someone explain it to me?

thanks!

> I don't understand V3(1.6v)

It is just symbolic. In real life, soultek would use a resistor-divider to set 1.6V from Q5 Q6 bases to V-.

It encourages Q5 and Q6 to pass 1V/500= 2mA of current up into the stages they feed.

Soultek-

What is this? A mike amp or a line amp? You fret about noise and say you have 44dB gain, but the schematic shows gain of 3.9 or about 12dB. 44dB would be fine for a mike amp but this is clearly not differential-input, usually a must-have feature in a mike amp. Reverse-taper pots are helpful in mike-level work (for noise issues you clearly understand), but not normally used in line amp where noise is not the biggest issue and it is often handy to be able to reduce gain to zero (silence). So I'm puzzled.

I -think- for lowest voltage noise you could run the input stage richer (though already you may be dominated by the 580 ohm resistance). Unlike BJTs, with FETs a richer operating point does not increase current noise.

The whole design is over-kill for my taste (which isn't a bad thing), except I wonder about second stage slew rate with those large MOSFETs and their high Gate capacitance being slapped around by a "mere" 2mA current. Too lazy to even estimate that.

What are you going to do with 25V peak signals? Is it all "headroom"?

Don't be too afraid of big electrolytics. They don't cost much, and the bigger they are the less they affect sound.

ok, thanks PRR, guess i'm just not used to reading these simulation schematics.

It looks an awful lot like a discrete op amp design to me. Like it could be a DI box, or mic pre, or any number of other things.

Okay, this is directly from my sim program so it's a little crude. Explainations:

-the 1.6V dc sources are these LEDs I use for biasing those transistors. They have a pretty consistent 1.6V drop across them. The part number is HLMP something I can dig up the info is someone is interested. These provide a reasonable temperature compensated biasing for the two NPNs. In the real circuit, I use one for each BJT to reduce the risk of any crosstalk path between the two stages. So the real circuit would have the anode connected to the base, the cathode connected to the negative rail and a 6k resistor between the base and ground.

-R5 can be a trim pot to dial in the standing current to compensate for differences in MOSFET threshold voltages.

-Note that the IR ouput FETs are beefy enough to drive a headphone provided you heatsink them enough.

-Note, that I could get a little more headroom, (and make the headroom more symmetrical) if I included another PNP stage complimentary to the Q6 NPN current source. (ie. connect the collectors of Q3 and Q6 and put a PNP current source at the collector of Q4. I had problems getting the biasing right, so I left it alone.

-There is about 800uA thorugh each JFET, this could perhaps be better optimized for noise, though I just messed around with this to get a bias that seemed to promote good operation.

-I found using R51 and C18 a much better way to compensate the opamp. Using miller compensation compromised the slew rate more than this method.

-OH! and the JFET's should be a 2SK389V, just simulator didn't have a good model for the dual part so i used the 170's. The BJT's are high beta Japanese Toshiba parts. Really nifty (beta=200-400). I try to use the GR coded ones. These can be bought from the spot market in quantity. I have been able to still find the JFET's although Toshiba has said they have stopped making them.

-I measured a slew of between 15-25uV/s

-Distortion was consistently less than 0.003, and less than 0.006 under heavy load and high gain.

-BW measurements are pending, couldn't accurately do with the AP. At highest gain (44dB) it is probably 10Hz-130kHz worst case. This can stand to be improved as well. The open loop gain is like 108dB and BW is I think 6-10Mhz. Increasing this (and restabilizing things) can improve my closed loop bandwidths. This can be possibly tweaked by changing the JFET drain current.

I donno, let me know what you guys think.

PRR, I was hoping for feedback from you, thanks.

The whole thing with this design was that I lacked studio experience to know what specs and features mattered. So I was unable to judge overkill/underkill early on. I picked a few design challenges and went after them. At this point it will be refinement. I will better understand the nuances of the design. The point this time around was to get something to work well.

As far as your comments:

-I first set out to design an opamp block that could be used for different applications (having a bunch of projects I'd like to tackle in the future). The first project I wanted to build was a DI with variable gain similar to a product Avalon makes. I desired to beat the specs they had published and add a transformer output stage as well. 44dB of gain is overkill for a line amp, but possible. I had 12 positions on the rotary switch so I went for it. R37 represents the rotary switch and is a 580ohm resistor which is it's lowest setting. Furthermore the min. gain setting of 12dB is meant provide gain to go from -10dBV to +4dBu levels, which is one of the primary purposes of this circuit. I had planned on using this with, say, a keyboard type application.

-I agree with the richer input stage, I haven't messed around with this as much as I'd like to.

-I chose those FETs because they had the least gate capacitance for FETs of that power rating. Also it would just be the gate-drain capacitance (still significant) that would be the issue, the gate-source capacitance would be bootstrapped out considering the configuration the device is run in (also provided the load is not crazy). But I agree, that would be a potential issuse. I admit I hadn't looked into this yet, thanks for highlighting it for me. The datasheet claims an input capacitance of 180pf, like you said, i'm too lazy to work out the numbers right now..

-The point of the 25V peak signals was sort of a leftover from an older idea that is more appropriate for another thread. I'll just say that I wanted significant headroom over the max signal level of the output transformer.

-Thanks for the advice on the big electrolytics, which was my next question..



-

I know this is always a raging debate. But I noticed some designs painstakingly design around having to use 'lytics. I started this way too, desiring to make a DC amplifier with a circuit to servo the offset back to nothing. But then I saw a bunch of well respected and otherwise good designs liberally use them, even though other companies even advertised "no electrolytic caps in audio path". I also saw the growing number of 'boutique' electrolytic caps out there as well. I know there is an associated distortion and leakage issue, but WTF?

So I just couldn't resist using the big cap in shunt resistance path of the gain control, it makes everything so nice from a design point of view. Then I began to think, if I use one there, why not just slap one on the output and forget about the servo.. I became convinced that I'd have to hear the difference myself before I'd own up to designing around using 'lytics. I just think it's creeping into audiophile territory...


Thoughts??


Then there's the other caps. Now using high quality caps here makes sense due to temperature stability and other real issues, but what about exotic materials? I hear you can get polystyrene's again now? those are supposed to be great but unavailable for a long time. Silver Mica? I think panasonic makes these now.. Then many reputed designers swear by a good 'ol ceramic cap. I know that SMD caps are still not as good as most of their quality through hole equivalents, but they are getting there too..

Thoughts??

also,

I also chose high headroom (which I admit is overkill for line amp applications) for the opamp so that I can use it for other applications where this would be more important. EQ circuits, mic pres, headphone amps, tape machine driver amps, and there's even some justification to use it in a phono pre too.

still might be overkill though i guess