Welcome to Solid State Guitar Amp Forum | DIY Guitar Amplifiers. Please login or sign up.

March 28, 2024, 04:23:14 PM

Login with username, password and session length

Recent Posts

 

Lowering noise in bipolar powered FET preamp

Started by sim0n, December 12, 2013, 09:29:05 AM

Previous topic - Next topic

sim0n

I've built myself the green channel (clean) of the famed Randall RG preamp to along with my TDA2050 power amp. It more or less the same thing as the famous red channel but with a treble bleed capacitor on the gain pot and no clipping diodes. I did however make some adjustments based what components I had and the amp it was going into:

  • moved the master volume to after the tone stack
  • made the entire thing run off bipolar 12V power (so that slightly changed the layout of the biasing)
My schematic is in the attachment. Part values are as the original's (heres a redrawn schematic of the dirty channel http://www.diystompboxes.com/smfforum/index.php?topic=79923.0 ),  33kohms on the drain, 2.2kohms on the source, I used 4.7kohms on source of both of the buffer FETs. A 100k gain control and a 100k volume control. My power supply is a 7812 and a 7912 linear regulator.

Alltogether it sounds pretty nice, a nice JFET overdrivey sound. The issue I have is that I suspect its more noisy than it should be, possibly due to the changes I made in regard to what reference voltages are.

With the input grounded, If I set the volume to 0 its dead silent, if I put my ear up to the speaker I can barely hear 50hz hum.
If I set the gain to 0 and thus effectively mute the first stage, theres a waterfall of noise coming from the second gain stage alone whenever volume is past 50% (where it needs to be for full clean volume if I set the gain control to around 25% for the cleanest of cleans with my hot pickups  xP ).
If I put a mic in front of the speaker and watch the VU meter in my computer I can measure that each stage is adding about 12dB of white noise.
To be fair, this is mainly evident when dimed, and with signal applied from 2/3 master volume and up the power amp is already clipping but still, lower noise would be better tjomg.

Is this just a fact of life with preamps of such design? if I were after a quieter amp I should have used something you're actually adjusting the gain of the stage and not just attenuating the signal afterwards? Use clipping diodes instead of clipping the transistors themselves so you don't have to amplify the hell out of your signal to get clipping?

Roly

This is a curious problem since JFET's are one of the lowest noise devices, and guitar levels are normally very much higher than the expected noise contribution of the device.

As a general rule the noise performance of an amplifier is defined by the first stage, and by its input resistance.  With guitar amps we are stuck with the need to present at least 1 Meg to the guitar, but the noise contribution of a FET stage is normally of the same order as the thermal noise in this first resistor, and of the order of microvolts rather than hundreds of millivolts, so normally only of concern with microphone preamps.  In this case we often see input devices paralleled, and paralleling four FET's should reduce the noise contribution of the first stage by half.

Another matter is the bandwidth of this first stage; the wider the bandwidth the greater the noise, and there is no advantage in having a first stage that has a greater bandwidth than the signals of interest, in our case about 5kHz.

At the moment your first stage bandwidth is limited by C2 in shunt across the input.  If you were instead to use a (much smaller) capacitance as a Miller capacitance between Drain and Gate you would be applying high frequency negative feedback and thus reducing high frequency noise.  Similarly...

Quote from: sim0nI should have used something you're actually adjusting the gain of the stage and not just attenuating the signal afterwards?

This would certainly help, however your first stage Source resistor isn't bypassed so it's a bit difficult to see how you could apply this idea.

I find your problem curious because this is normally not a problem with a JFET input stage at guitaring levels.  You may find some advantage in changing the gain distribution between stages; making the first stage the high gain stage by bypassing its Source (and incidentally allowing gain reduction by local degeneration, rather than gain then attenuation) and not bypassing the Source of the second stage.

This current arrangement appears to fly in the face of the normal idea of putting the gain up front to get the signal up out of the noise; while here we have the gain following the first, noise generating, stage.

Quote from: sim0ntheres a waterfall of noise coming from the second gain stage alone whenever volume is past 50%

I find this surprising and unexpected, but ATM can't see any good reason why this should be so.  Since you are taking measurements I'd be inclined to increase the current through this second stage, say by a factor of four, and see what that does to the noise.

HTH
If you say theory and practice don't agree you haven't applied enough theory.

Loudthud

Without a bypass capacitor on the first stage, it's voltage gain has to be something less than 1/2. This is because the voltage across the Source resistor is more than twice as much as the Drain resistor. The current is the same so the resistance is proportional to the voltage drop. You would be better off just using a source follower. The advantage of the circuit is that the Source resistor sets the Drain current fairly accurately so you don't need a tweek.

With only 12V rails and the first stage only able to swing from the plus rail to a couple of volts, you want a gain of about 2 to mimic the headroom of a tube stage. A bypass cap with a series resistor would increase and control the gain.

J M Fahey

1) if you want the Randall sound, build it original and with 24V .
You will not only lose headroom but *gain*, gobs of it.
FETs are incredibly sensitive to +V voltage and all normal ones are wimpy with 9V, period.
As Loudthud pointed, typical gains are 2X per stage, *even if source bypassed*, because you have an *internal* resistance you can't bypass, simple as that.

2) your hiss waterfall is not normal, so ot's either RF oscillating (you can't hear anything above 20KHz, but its interference with normal sopund, they sound similar) or you wired your FETs upside down.

In theory they are symmetrical, you can interchange Drain and Source, but in practice one configuration hisses like H*ll.

3) dig around for the proper 24V PCB layout, build it, feed it 24V and smile.
Oh !!! forget those abominations known as drain trimmers, use proper fixed resistors , 33K or whatever, and try different Fets until voltages are normal, don't try to "correct" a poor/weak/bad one with a trimmer.

If you need 3 Fets, buy a dozen and handpick the best.

4) and if you want good clean sound, build Rodd Elliott's guitar preamp, designed around dependable Op Amps.
Killer fendery sounds, similar to old Yamaha and Roland amps.

teemuk

Quoteif you want the Randall sound, build it original and with 24V .

This.

And the "original" clause also includes those clipping diodes and the proper feedback scheme (both voltage and current) in the power amp section.


sim0n

#5
Alright, cleaned enough bench space to hook things up to take some measurements. Here is the schematic of what I've actually built, with actual component values and measured DC voltages. I tweaked a couple of component values while I had it built on the breadboard, so I attributed the noise at that stage to the fact it was built on a breadboard.



Hooked up to my oscilloscope, with a 1khz test signal from a little test oscillator, the first stage has a gain of about 9x. My oscillator doesn't have enough output to drive it into clipping but at 14v peak to peak theres some onset of distortion (I moved the source capacitor to the second stage because my guitar's pickups have enough output to clip this stage  :cheesy: )
The second stage has a gain of 37X and goes finishes in all out asymmetric relatively soft clipping with 20 volts peak to peak.

Noise-wise theres about 1mV on the power rails (and it seems to be evenly distributed over the spectrum). On the gate of the first stage with the input grounded about 1mV, on the first stage's drain maybe 2 or 3mV (measured as peak to peak voltage). With the gain set to 0% theres maybe 1.5mV on the second stage's gate and 15mV of the drain. With the gain at 100% theres 50mV p-p of noise at that point and it follows through in the rest of the circuit.


So like I mentioned, the preamp functions and sounds fine.  And with both volume and gain at 100% there is way too much output on tap as the power amp chip already clipping with and input 2v peak to peak and even after the tonestack losses theres still way too much. But I still feel the total amount noise should be lower (50mV of noise x 40x gain of the power amp = 2V on the speaker  :P)

For now I'm suspecting somethings going on with the gate voltages. As if they're not at the potential of the lower rail it would imply that theres current flowing into the gate, which can't be right for FETs  :o

edit: hmm, or my multimeter's internal resistance is pulling the gates up because I have it referenced to 0V and not -12V

J M Fahey

Quoteedit: hmm, or my multimeter's internal resistance is pulling the gates up because I have it referenced to 0V and not -12V
That's what's happening, glad you noticed it.
Onstead of a +/-12V supply, reconnect it as a 0/+24V one and you'll be like the original one.
And see that your Fet gain stage has only 9x gain (compared to some 50X in a similar tube triode stage)
With +9V supply and those pesky drain trimmers favored by some "designers" (won't give names  ::) ), gain is useless 2X or thereabouts.
With a 25/30V supply I regularly get 20X gain per stage, which is usable to emulate tube preamps.

Roly

Quote from: sim0nI'm suspecting somethings going on with the gate voltages.

Agreed.  I had an intuitive feeling that the Gate of the first FET was forward biased, 'tho I couldn't and can't see any obvious reason why that should be, but the voltages you give confirm it, and that will give you a s&@tload of noise.

The reason for going for max gain in the first stage is the same for audio as it is for  VHF - it gets the desired signal up out of the noise.
If you say theory and practice don't agree you haven't applied enough theory.

teemuk

#8
QuoteThe second stage has a gain of 37X and goes finishes in all out asymmetric relatively soft clipping with 20 volts peak to peak.

You do realize that inclusion of the diodes present in original Randall circuit would reduce that to about 8Vp-p or 1.2Vp-p depending on overdrive setting....

Quote...the power amp chip already clipping with and input 2v peak to peak and even after the tonestack losses theres still way too much.

I wonder why...

sim0n

The Randall had a clean channel too, without any diodes  :P


I've since managed to fry the power amp chip by accidentally shorting something with a multimeter probe while I trying to measure something so this project is paused until another TDA2050 arrives