Acceptable amount of distortion in a clean preamp? - fetzer

[blackieNYC]

What oscillator voltage should be used to tinker with a clean preamp?  A guitar output with buckers can be 100mv to as much as 1 volt?  Pedal owners are a forgiving bunch, but what is an acceptable THD for a "clean" guitar Pedal circuit? I'm feeding a sine wave at -10db (350mv)into a fetzer circuit that I believe to be built correctly via runoffgroove.  I'm getting 4% thd with 10 dB of gain with -10 in. (And, I'm 3 dB down at 10 k, and the article states this should be higher).   So, what's an acceptable level of distortion in a clean pedal? I'd be building my pedal just by ear, but I ran into some questions that had to be answered by test equipment.

[R.G.]

Scope a wide-open neck humbucker chord hit with a windmill pick attack.

If you want clean with all guitars, that peak is your target.

In general guitar owners are oblivious to THD; many don't know what that is.

You want clean, make it clean. You can get well under 0.1% THD with a simple opamp and some power supply work. You want high gain AND clean, build a bigger power supply.

[Mark Hammer]

While I can understand the desire to have a pristine clean signal, once it leaves the preamp, it still has a long way to go between there and your ears.  It still has to pass through a power amp, speakers, and maybe even an output transformer.  And all of those will add their little touch of harmonic icing to the cake.

That is not to say one should not strive for clean up front, but it is to say that the average guitar rig is likely to still dirty things up a bit, so don't drive yourself crazy about nailing the lowest possible harmonic distortion at the front end.

Apart from that, nothing wrong with upping the supply voltage, socketing the JFET, and trying a few different ones out, just to see if it makes a difference.

Do keep in mind that the Fetzer is intended to be a solid-state implementation of a Fender front end.  While it is not intended to deliver Carlos Santana overdrive, much like the amps that inspired it, neither was it intended to attain audiophile quality.  What Uncle Leo considered "clean" may have simply been an acceptable amount of THD.

[blackieNYC]

Thanks for the replies, gentlemen. I don't need much gain, I'm building a splitter/looper/mixer thing with one dry and two loops, each with a buffer, and a mixer with volume for all three. All with just 4 FETs.  And two more in case a phase flip is needed. It's going to be cool. Anyway, that's good to know about the fetzer. You convinced me to use a Tillman preamp, the output of which splits to the two buffers for the loops and straight to the mixer with the dry.  But, I just measured a Tillman I recently built and its giving me 4% thd at -10db in.  Is minus 10 (250 mf, rms I think) too much level for testing.  Or is 4% acceptable distortion?  I wouldn't expect the Tillman to do this.  I honestly have to stick with 9 vdc, but I should be able to do this cleanly, no?

[blackieNYC]

250mv, that is.  -10db

[R.G.]

Or is 4% acceptable distortion?  I wouldn't expect the Tillman to do this.  I honestly have to stick with 9 vdc, but I should be able to do this cleanly, no?

(1) For guitarists, 4% distortion is just fine as long as it's the right distortion. Guitarists really like the soft even order distortion of single ended jfets and tubes. Really, musicians in general like up to a few percent of asymmetrical and particularly second order distortion. It is perceived as a subtle sweetening of the sound, and musicians like that. Hifi tweakos go insane over it as "inaccurate". One man's ceiling is another man's floor.
(2) The Tillman is a single ended JFET. 4% is a little high; I'd have expected more like 2%, but it's not out of the question. Without feedback, single device distortion in this range is to be expected.
(3) With 9V as a limit, no. The proximate cause you're having is that the incremental gain of a JFET (... or bipolar, or tube, or MOSFET) varies with its channel current. Single digits distortion is really very good for things other than vacuum triodes, which are remarkably linear for what they do. But then they have a lot of volts to do it with. The larger the proportion of the available voltage you eat up, the more distortion you'll have in the case without negative feedback.

And "clean" is relative. There was a time when 4% was pretty clean, and 1% was what hifi people advertised as GREAT.

[sjconolly]

FWIW which may not be much...

I haven't tried building the Tillman circuit since I can't seem to come up with the right type of jfet locally, but out of curiosity I put the circuit into LTSpice and came up with 6% distortion, so 4% wouldn't surprise in the real world. Using 18 volts instead of 9 will probably help a lot.

Because the guitar string produces a wide and prominent series of harmonics, I believe the harmonic distortion is a lot less noticeable. 2nd harmonic is fine, so I prefer to look at the level of the 3rd harmonic in DB below the signal. If it's 60 db down from the signal it's plenty clean, if it's 30 db down I have problem.

That said, I've done a couple of opamp circuits now that are quite clean, but sound bland compared to a simple transistor circuit with quite a bit more THD.

[GibsonGM]

That said, I've done a couple of opamp circuits now that are quite clean, but sound bland compared to a simple transistor circuit with quite a bit more THD.

I'd keep that in my thoughts as well...a completely clean (pre)amp is likely to sound quite flat.  That is, unless what you're feeding it ALREADY has the right amount of THD for your ears.   

This is why way back when, plugging a guitar into a hifi rig, with hifi speakers, WORKED but sounded like junk.  Lifeless.

[blackieNYC]

Thanks all!  I'm starting to realize that with just 9v, the common source circuits (drain output) are the only ones giving me gain, and always 4.5-6% thd.  The source output FET circuits are not capable of gain, you'll lose a dB, but are clean as a whistle.  Oh, and the drain output inverts, so I have to attenuate, and the phase inversion will contribute its 5 % distortion.  But I'm still going to do this giant pedal with FETs, even if it kills me. So I'm using Orman's splitter and losing a dB but keeping it clean, and summing the fx loops together with a fetzer.  I would have liked to hit the pedals a little harder, but they'll receive a buffered feed, and I'm just going to have one gain stage (to keep the thd reasonable), so that will be a fetzer on the output.

[R.G.]

If you're not completely hung up on using ONLY FETs, consider this:  http://geofex.com/FX_images/Onboard_Preamp.pdf
It's done with bipolars, but I believe I could replace the input bipolar with a JFET and tweak it in.

The output is nearly rail to rail, *remarkably* clean, and has frequency response from DC to daylight - well, OK, a few hundred megahertz. It's input is about 1.5M, and could be bootstrapped to be even higher. It can be made to have gain by dinking with the output a bit. It can be switched to work from any power supply voltage up to about 30V, I believe. This one was optimized for clean, non-loading buffering with really, really small current drain from the power supply.

[PRR]

4% 2nd-order distortion on BIG signal through a SINGLE-stage is fine.

I had a tube-amp, the "distortion" channel could be dialed dead-clean (to diode-clipped) and the "clean" input was classic Fender with 5% 2nd harmonic on big inputs. I have pretty bad ears, and have trained for "clean" (reproduction systems), but I liked the simple 5% 2nd harmonic much more than the "perfect" input. It's not "dirty", it is "richer".

Multi-stage systems, if "all" stages distort, get you into IM effects which "haze-up" the music. That's why a studio console and recorder with 567 stages to go through must show 0.01% per stage.

> -10db (350mv)into

What OUTput?

If the stage has gain of 5, then 0.3V in (-10 what... dBv? dBu? dBm??) is 1.5V output. That's a big guitar signal! Also a rule-o-thumb is that a vacuum triode (FET similar) can make peak out voltage 20% of supply voltage and do about 5% THD there. 20% of 9V is 1.8V peak, 1.2Vrms. If you get 4% at 1.5Vrms out, you are doing very good, for a simple 1-device amplifier on 9V supply.

> I don't need much gain

Put a number on that. Gain of 1.1 just so it don't sound like "less"? Gain of 2 to cover a mixing loss? Gain of 10 to SMACK a lazy/hissy efect?

Reducing gain may do two things. First your output will be a smaller fraction of your 9V supply, so it doesn't works as hard and is lower on the THD curve. Second, many smart ways to reduce gain also improve linearity (lower THD).

The 10-cent ultimate is a TL072. You can get the exact gain you need, and the 1000:1 of excess gain cancels distortion down to the 0.01% level. '072 also allows high input impedance, will drive any load that should be found in a gitar chain, works well on 9V (less-good as battery sags below 7V, but everybody has a power-pack?), and you can buy a box-full for a penny a leg.

R.G.'s buffer is another path.

[brett]

Hi
As long as it's not a non-musical sound like breaking glass, I can barely hear 10% distortion. Probably more than 10% for octave and 2 octave up. People have posted clips of a clean and a few % distortion of 'musical' distortion. From memory, the crowd had 50% accuracy in identifying the distortion.
Oh yeah, never play through a speaker. They add heaps of distortion. True! 
cheers

[blackieNYC]

It's a fx looper with buffered loops. clean , no gain- or almost a dB in loss on these sends.  They return into 10kpots, and 100K Rs sum them into a fetzer with a 201.  That has enough gain to make up for the mixer loss (at this point I'm a little above unity), and then I've got a foot switch bring in a big source R HP cap, which yields about 6dB in boost - post everything.  Haven't actually listened to that boost yet. Supposed to deviate the fetzer from true tube mimicry. Whatev. Hope it sounds nice.
Built a Tillman. Sounds nice. 5% thd.  Didn't notice.
DBu of course. As measured by the $10,000 Audio Precision I borrowed from work to make dirt pedals with!  Principle! That's why I'm going FET, because the TLO72 has 3 FETs and 13 BPTs - and I'm certainly not going use THAT to drive my pedals (a string of 152 TL072s, I know ).
Just about 3 threads earlier, RG and others have put up more FET selection details I'm making use of.  Time to drill some holes.

[blackieNYC]

R.G. 's FET design tips about the same time this one got started
Thanks
http://www.diystompboxes.com/smfforum/index.php?topic=104341.0

[blackieNYC]

Love to see the FET version of your onboard buffer.  And 6 volt coins can do it?  Cool. I was going to chop some AA battery holders in half and stuff a 544 camera battery in there. The 2032 is actually higher mAH
Brett and Paul also offer great info in the aforementioned thread.  Thanks all.

[R.G.]

I'll go mess with it a bit for a JFET front end.

I actually went to the discrete design because the opamp solutions I was looking at would have had either more current drain or more noise, and I was after something which ran for a Long Time (tm) inside a guitar.

For a situation where you have all the power you want, it's really difficult to beat an opamp buffer. It can be done, but designs tend to get complex fast. The reasons for going discrete tend to be either lower power, lower noise, higher bandwidth, lower fuzzy concepts from hifi tweako land, or advertising.

In general, the best, most linear single-device amplifiers have a few percent distortion if they have not much feedback. Negative feedback uses excess gain to suppress distortion. Note that an emitter resistor or source resistor is negative feedback, and this is why followers have low distortion, and why unbypassed emitter/source amplifiers have lower distortion and lower gain. The higher the open-loop gain to closed-loop gain ratio, the more the normal open loop distortion of the amplifier inside is suppressed.

My onboard buffer is a high(ish) gain, high feedback design. The first device has voltage gain to the collector, and this is fed directly into the input of the second device, which multiplies by its gain. This is then fed back into the emitter of the first device, which acts like an inverting input in an opamp. With 100% of the output signal going into the feedback, all of the forward gain is used to suppress internal distortion from being seen at the output. The difference is that the forward/openloop gain is smaller than an opamp, although a lot bigger than a single device.

Trading out the first device for a JFET will increase the input impedance of the internal amplifier stage somewhat, and lower the forward gain, as the transconductance of a JFET is lower than that of a bipolar.

[R.G.]

I did some quick messing with the simulator.

You can trade in a J201 with no other changes, because of it's anomalously low Vgs needs. You can put in a low Vgsoff device like the 2N5485, 2SK30A, 2N5292, etc. by only changing one resistor, the lower resistor in the biasing string. These seem to bias well enough changing the 1.7M resistor on the bottom to 470K.

Doing this primarily changes the distortion. In the NPN version, distortion with smallish signals that don't get near the edges of the power supply range on the output seem to have about 0.2% THD, with the second and third being a lot of it, but long trails of distortion products much further down.  Subbing in a JFET gets you up to about 0.45 to 0.6% THD, and the distortion profile is similar.

Thinking about it, we would expect the THD harmonic profile to be similar, because the input device is being run with very small signals indeed. Most of the distortion would come from the signal swing in the second device, which doesn't change all that much. I believe the changes that are there are from the lower open loop gain of the thing with a JFET front end.

This is crude work, of course, but it offers some directions.  Putting a JFET front end into that circuit mostly bucks up the input impedance and gives a little more THD, but without making the THD profile be the big, fat second harmonic that adds subjective sweetness. Not unexpected.

If you want sweetening, run a JFET open loop. Or a triode. These cases will need help in not being loaded, so you'll need a buffer, possibly something like a JFET or MOSFET input opamp, after them to take the load off. Otherwise, the loading will mess with your gain and sweetening device.

If you want clean, skip the JFET and just use the opamp.

If you want bragging rights, use a discrete amplifier with special hoopla designs. Just putting in a JRE990 often does that, but it amounts to an opamp.

[blackieNYC]

Arghh!  It's all gone to hell.  The AMZ splitter is fine, common drain so the gain after each buffer is down a dB. If I use a fetzer as the summing amp for the dry, fxA, fxB sum, it distorts. The signal at the summing point ( fetzer gate) is low, but I can't clean this up. I tried to keep it low there, and add another gain stage after it, with an attenuator before it, to keep it from clipping, but if I use any means to brin the signal back to unity, it (guitar with hb's, not tone) it's got too much crunch. And I was hoping the 2nd output stage would give me a boost.  And, if I build the aaroncake mixer with the original source-follower summing FET, the signal output with one channel maxed is down 10 dB!  Look at this, i think it requires a serious boost: http://www.aaroncake.net/circuits/mixer1.asp
I don't want to throw on the towel on a 9volt FET mixer and go op amp, but I'm starting to crack up. RG - I haven't tried the FET version of the booster you mentioned, partly because I don't think I have a handle on the changes you propose for FETs, and partly because it is huge and I'm running low on breadboard!  Is it going to give me the clean gain?
Guys, help me off the ledge!  In short I wish to cleanly sum a 3 ch mixer to unity and then have around 10dB of boost available.  I want to take one more crack at all-FET.  When the time comes for op amps, with my head hung low, I would think that I'll have less trouble summing with gain.