Mike Sulzer Posted January 5, 2006 Report Share Posted January 5, 2006 Well, I changed that source resistor to 22K. I can hear a very subtle change in the mids/high mids as a result. It probably would be much more noticeable with a 20' or longer cable. Better safe then sorry, I guess. ← You are probably hearing a difference in the small amount of distortion produced by the circuit, not its linear freqency response. With the 22k resistor, you have changed the operating point of the FET. Quote Link to comment Share on other sites More sharing options...
Paul Marossy Posted January 5, 2006 Author Report Share Posted January 5, 2006 With the 22k resistor, you have changed the operating point of the FET. How so? Quote Link to comment Share on other sites More sharing options...
Mike Sulzer Posted January 5, 2006 Report Share Posted January 5, 2006 With the 22k resistor, you have changed the operating point of the FET. How so? ← From the specs, with 22K in the source, you have a drain current of about .1 ma, or about 2-2.5 volts across this resistor. At low frequencies you can swing about 2 volts in the negative direction, not ideal, but more than you need. Driving the 40K load of the cable cap at 5 KHz reduces this swing somewhat, but not too much. Remember the curves for FET; as you lower the drain current, a given increment in VGS has less effect in further lowering the drain current. This is the essential non-linearity of the FET. So with 200K you probably get about 4 or 5 volts or about .02 ma, but the specs do not go this low, so I am not sure. To drive the same voltage and current into the cable cap requires a five times larger relative* current change than with the 22K. Thus the non-liearity is greater. *That is, relative to the no signal operating current. Quote Link to comment Share on other sites More sharing options...
lovekraft Posted January 5, 2006 Report Share Posted January 5, 2006 Mike, how does changing the operating point of a source follower change its tonality? Other than possibly having it clip on one half-cycle, what audible differences are the result of the operating point? According to SPICE, the DC operating point falls from about 1.72volts (with a 220K source resistor) to about 1.56 volts ( with a 22K), and my breadboard with a random 2N5457 seems to bear that out (actually about 1.75v vs 1.55v with a 9.1 volt battery). Surely that's not enough difference to maker an audible change (provided the circuit isn't clipping, fo course). Frankly, with either of those operating points, I don't see how you can expect to get much more than 1 volt P-P unclipped out of it at any frequency. Quote Link to comment Share on other sites More sharing options...
weaponepsilon Posted January 5, 2006 Report Share Posted January 5, 2006 Hey. This is a great idea since my Upright Piezo screams like I'm going to kill it. I only have one problem. When I click on the PDF file of the layout, it tells me the file is damaged and doesn't load. Could you send it to me or somehow fix it/make it public? Thanks. Quote Link to comment Share on other sites More sharing options...
Paul Marossy Posted January 5, 2006 Author Report Share Posted January 5, 2006 (edited) Hey. This is a great idea since my Upright Piezo screams like I'm going to kill it. I only have one problem. When I click on the PDF file of the layout, it tells me the file is damaged and doesn't load. Could you send it to me or somehow fix it/make it public? Thanks. weaponepsilon- It opens up fine for me... did you recently upgrade to Acrobat 7.0? I did and I get that error message sometimes even though I know the file is perfectly good. If it's still a problem for you, I can email it to you if you give me your email address. Edited January 5, 2006 by Paul Marossy Quote Link to comment Share on other sites More sharing options...
Mike Sulzer Posted January 6, 2006 Report Share Posted January 6, 2006 Mike, how does changing the operating point of a source follower change its tonality? Other than possibly having it clip on one half-cycle, what audible differences are the result of the operating point? According to SPICE, the DC operating point falls from about 1.72volts (with a 220K source resistor) to about 1.56 volts ( with a 22K), and my breadboard with a random 2N5457 seems to bear that out (actually about 1.75v vs 1.55v with a 9.1 volt battery). Surely that's not enough difference to maker an audible change (provided the circuit isn't clipping, fo course). Frankly, with either of those operating points, I don't see how you can expect to get much more than 1 volt P-P unclipped out of it at any frequency. ← OK, this is a bit more extreme than I thought. The drain current differs by almoat a factor of ten in the two cases. The obvious difference is the clipping level when driving a 40K load. With the larger resistor, you are starved for current, and so you can swing only about .3 volts in the negative direction. With the higher curent, you can get closer to the 1.5 volts. The more subtle difference occurs short of clipping. The same magnitude of change in VGS in the positive and negative directions does not give the same change in drain current. If we make VGS more negative with a set of equal increments in voltage, each additional increment results in less decrease in drain current than the previous one. This is non-linear behavior. The farther you move off the operating point, the greater the effect of the non-linearity. Suppose in the lower current case and the 40 K ohm load, we drive with signal that requires the FET to nearly turn off. The non-linearity is large. Suppose we go to the high current case, same signal voltage. Now the FET does not get anywhere close to turning off, and the non-liinearity is smaller. Quote Link to comment Share on other sites More sharing options...
Paul Marossy Posted January 6, 2006 Author Report Share Posted January 6, 2006 I used it some more this morning, and I think it is better with the 22K source resistor - it seems to have mo "umph". Quote Link to comment Share on other sites More sharing options...
lovekraft Posted January 6, 2006 Report Share Posted January 6, 2006 OK, just so we're all talking about the same circuit, I drew it up, with the cable capacitance and the amp's input impedance added in. As shown, with a 1.5volt p-p input signal, I'm showing a THD of just above 1%, almost all desirable 2nd harmonic ("tube warmth"), and a high frequency -3dB point of about 17KHz. Since guitar speakers start to roll off heavily around 6KHz, and most of the signal content of an acoustic guitar above 10KHz is fret clatter and finger noise, this looks like a fairly ideal solution, provided noise levels are low enough, and a 2N5457 working at what is basically unity gain shouldn't make enough noise to be audible. If you need to drive a 600 ohm line level input, you'll need a good DI box, but for use with guitar amps, this looks more than adequate. Using a 220K source resistor appears to raise the distortion (ca. 2%) and lower the -3dB rolloff (12K), so unless you need to drive a much heavier load, I see little point in using the larger source resistor. Quote Link to comment Share on other sites More sharing options...
Paul Marossy Posted January 6, 2006 Author Report Share Posted January 6, 2006 Thanks for the simulation lovekraft! The presence of 2nd order harmonics would explain why I feel it sounds warm. The frequency response seems to be pretty good, too, as I can hear some finger noise and stuff - but not enough to be annoying. Like I said, it's a great little circuit! Quote Link to comment Share on other sites More sharing options...
JohnH Posted January 6, 2006 Report Share Posted January 6, 2006 Well this has become a very useful thread. I was interested in how the frequency response turned out to be much more robust than I thought. I also got the same results in a pSpice simulation. I used a 2N3819, which has a higher VGS than the 2N5457, and sure enough, the source voltage rose to about 2.9V. i wonder if this would give more headroom? Lovecraft - what is your view on the role of the 1pf and 4.7mfd capacitors in your posting? Quote Link to comment Share on other sites More sharing options...
Mike Sulzer Posted January 6, 2006 Report Share Posted January 6, 2006 OK, just so we're all talking about the same circuit, I drew it up, with the cable capacitance and the amp's input impedance added in. As shown, with a 1.5volt p-p input signal, I'm showing a THD of just above 1%, almost all desirable 2nd harmonic ("tube warmth"), and a high frequency -3dB point of about 17KHz. Since guitar speakers start to roll off heavily around 6KHz, and most of the signal content of an acoustic guitar above 10KHz is fret clatter and finger noise, this looks like a fairly ideal solution, provided noise levels are low enough, and a 2N5457 working at what is basically unity gain shouldn't make enough noise to be audible. If you need to drive a 600 ohm line level input, you'll need a good DI box, but for use with guitar amps, this looks more than adequate. Using a 220K source resistor appears to raise the distortion (ca. 2%) and lower the -3dB rolloff (12K), so unless you need to drive a much heavier load, I see little point in using the larger source resistor. ← Very interesting, Lovekraft. I learned electronics before there was Spice, but I guess I should get a copy, Did you measure the distortion at 500 Hz, or 5000 Hz? This could make difference. I think the only reason for using the larger source resistor would be to get nearly infinite battery life. By the way, the 1 pf cap probably does not do a lot for rf suppression without a choke in the signal line before it. But I guess it cannot hurt. The 4.7 microf cap is power supply bypass, always a good idea, but I bet it work fine without it. I think that you could get lower distortion (if anyone wants it) with this FET by removing the 10 Mohm resistor from ground and connecting to a voltage divider coming from the battery and providing about 2 or 3 volts. Or use the FET JohnH did. Quote Link to comment Share on other sites More sharing options...
lovekraft Posted January 6, 2006 Report Share Posted January 6, 2006 Bypassing the power supply with a large cap ( I've used up to 100uF when it was all I had) is just good practice, especially with battery power supplies - I do it without thinking about it unless I can't afford the board space, and I don't make tiny things very often any more. The 1pF cap is for RF bypass, I assume, and while it's probably not as critical here as it would be in a high gain circuit, it's cheap insurance - Justin Wilson used to say he was trained as a safety engineer, and that's why he wore both a belt and suspenders, and I'm all for that attitude. It should shunt that pirate radio station out of your signal chain, keeping those "Spinal Tap" moments from happening. I think the reason we're seeing a much higher rolloff is because, as Mike Sulzer implied, the output impedance is actually a product of the JFET's source resistance in parallel with the source resistor - I sure wish I had paid more attention to those physics courses I had to take in college! I'm always amazed by these guys who can derive approximations for gain, frequency response and equivalent noise figures for discrete circuits with a slide rule - it's like magic!! Quote Link to comment Share on other sites More sharing options...
Paul Marossy Posted January 7, 2006 Author Report Share Posted January 7, 2006 Just for the record, I used a 20pF cap instead of a 1pF cap. I felt that 1pF was just too small... Quote Link to comment Share on other sites More sharing options...
lovekraft Posted January 7, 2006 Report Share Posted January 7, 2006 Paul, I agree that 1pF sounds just a bit small. I've been using 33pF ceramics with good results (mainly because I got a smokin' deal on 100 of 'em). Mike, using a voltage divider to bias the FET is a great idea, and it does apparently lower distortion and boost headroom, but it should also pretty much eliminate any noise from the bias resistor (classic "noiseless" biasing). It also (for reasons I don't understand, which you might be able to explain) raises the bandwidth just a touch. Here's my take on your suggestions for optimizing this circuit: I think that's about as far as we're gonna be able to improve this configuration with a 9 volt battery for power - current consumption is higher than the original, but you still ought to get at least a couple of hundred hours out of a fresh Duracell. Quote Link to comment Share on other sites More sharing options...
Paul Marossy Posted January 7, 2006 Author Report Share Posted January 7, 2006 Interesting. I think I might try this biasing arrangement on the next one that I build - more headroom is always better. So, just how much headroom are we talking about? Quote Link to comment Share on other sites More sharing options...
lovekraft Posted January 7, 2006 Report Share Posted January 7, 2006 Looks like under .25% THD with a 5KHz 2volt P-P sine input. and about 2%THD ca. 3.25volts P-P where it starts to clip significantly. With the input attenuator, it should handle pretty much anything we can throw at it. Quote Link to comment Share on other sites More sharing options...
Paul Marossy Posted January 7, 2006 Author Report Share Posted January 7, 2006 Well, 2V P2P should more than enough for just about anything with piezos. I think I will incorporate that into my next build(s). Quote Link to comment Share on other sites More sharing options...
Mike Sulzer Posted January 7, 2006 Report Share Posted January 7, 2006 Looks like under .25% THD with a 5KHz 2volt P-P sine input. and about 2%THD ca. 3.25volts P-P where it starts to clip significantly. With the input attenuator, it should handle pretty much anything we can throw at it. ← That's looking really good now. One comment on the reduced bandwidth with the large resistor and small drain current: The transconductance decreases as you lower the current. You can tell this is happening because the curves of drain-source voltage vs. drain current get closer together at lower current or more negative VGS. The output impedance of the source follower rises as the transcondance goes down, and so the bandwidth into a capacitive load decreases. One thing is still confusing me. With the original circuit with the 200K resistor, it looks to me like the output at 5000 Hz with a 1.5 v p-to-p waveform input would be very triangular, kind of like slew rate limiting with an op amp; that is, lots of distortion. Does Spice show this? You might have to put a large resistor in series with the input voltage generator. Quote Link to comment Share on other sites More sharing options...
JohnH Posted January 7, 2006 Report Share Posted January 7, 2006 (edited) This is looking good with that mid-voltage bias. Im guessing that with that arrangement, it is less sensitive to FET properties. two more thoughts: My understanding is that with JFETS, the gate gets biased to a lower voltage than the source. To set the source voltage to mid range 4.5V, ie half the supply voltage, to get maximum headroom, should R7 be less than R5? (This could then bias the gate to say 1.5v to 3v depending on the FET). In fact, why not make them both part of a trim pot, then you can wire it up and set the bias exactly for that particular JFET I am always looking for a built-in design where the battery lasts for ages. The 10k values for R5 and R7, although not taking much current, are still more than doubling the current consumption. I suspect that those r5 and r7 resistors could be changed to something much higher, say replace them with a trimpot of 1M wired from + to -, with the wiper feeding the 10M resistor. John Edited January 7, 2006 by JohnH Quote Link to comment Share on other sites More sharing options...
Paul Marossy Posted January 8, 2006 Author Report Share Posted January 8, 2006 I almost always use a trimpot on at least the source of FETs to compensate for their high variability. And, that way, I can dial them in exactly where I want them. Quote Link to comment Share on other sites More sharing options...
lovekraft Posted January 8, 2006 Report Share Posted January 8, 2006 Simply change the voltage divider to a pair of 10Megs (or a 22Meg trimmer, if you prefer), replace the 10Meg gate resistor with a jumper, and increase the source resistor back to 220K - that'll drop current consumption to about 30uA without compromising noise, distortion and treble response performance very badly. That should make the battery last, but you'll have to try it to see if it sounds OK. TANSTAAFL, so design decisions are always about compromising one set of specs to get better numbers elsewhere. Quote Link to comment Share on other sites More sharing options...
ROBERTLATHAM1 Posted January 10, 2006 Report Share Posted January 10, 2006 wow ok lost but i think i will try this thing out wow!!!!!!!!!! electronics to an electrician is like brain surgery to a mechanic Quote Link to comment Share on other sites More sharing options...
JohnH Posted January 13, 2006 Report Share Posted January 13, 2006 I have been playing with this circuit over the last couple of days, to pick a version to go into a new rewiring build. I wanted to explore the extremes of the various parameters in order to understand them. I'm using a 2N5484 JFET, and Im expecting it will perform similarly to the 2N 5457. I am biasing it with a 2.2M gate to ground and a 3.3M gate to positive. This sets the source voltage at about 4.7 V. I tried a wide range of values for that source resistor (discussed earlier as 220k or 22k). When feeding via a 20' cable into an amp, I found that even a 1M resistor sounded fine. I then tried it with a greater load on the output, representing perhaps some not very well designed effects box. I loaded the output with a 27k resistor to ground for this experiment. The result was, with a high source resistor, you get an increasing amount of low order harmonic distortion. Actually it did not sound too bad, but not what I was looking for. With the source resistor at about 150k or less, this was inaudible to my ears. So I have settled on a 100k source resistor, confident that this is a robust value that will deal cleanly with any practical load on the output. Current consumption is a nice small 50 microamps. The sound shows a good improvement in treble clarity compared to just passive, and no noticeable output level reduction. This buffer will also be the key to get my independednt volume controls in series and parallel to work. Thanks again for this thread. John Quote Link to comment Share on other sites More sharing options...
Paul Marossy Posted January 16, 2006 Author Report Share Posted January 16, 2006 The sound shows a good improvement in treble clarity compared to just passive, and no noticeable output level reduction. This buffer will also be the key to get my independent volume controls in series and parallel to work. Thanks again for this thread. Cool, thanks for your insights. And I'm glad that it was of some use to you as well. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.