Working preamp proto - clean up help

[lion]

I've never got into proper breadboarding. I just build - often reusing previous test builds or circuit parts, building in sections and with "whatever" parts/values I have at hand till I get it working - or not (admitted it usually turns out a mess in the process, and from the multiple changes as I experiment - but that's the way that suits me best )

Now I have this project working as I want, and it's time for clean up - before I consider building it properly and from scratch with correct/best components and values.



DC supply is +/-9V
At this point there's probably one or two DC blocking caps which can be omitted, and values way above what's needed. Probably the same with resistors (EDIT: I mean one or two might be "leftovers" and have no/little purpose as the circuit ended, voltage divider values might need scaling down etc). Freq response wise I basicly want it to pass "everything", as the actual voicing of the echoes is handled in the external digital delay (very accurate with multiple filters, and adjustable per individual patches).

With respect to pot values - I'm never sure what's the proper values for optimal loading/impedance/noise.
Looking at it as I type I'm wondering if the "echo vol" should be moved to between T2 and the final buffer - if it makes any difference at all.

Anyone care to take a look and advise - basicly on anything (odd, stupid, plain wrong etc) - in order to optimize the circuit for best practice and performance. It would also help me in my learning process to know what I've got right or wrong. Thanx.

Erik

[PRR]

You sure want a gate resistor on T1.

Separate bias, and cut-down gain-up, can be consolidated.

U1B is not essential, T2 can drive a 200K load.

Caps marked "?" should not be necessary.

[lion]

Thanks Paul, much appreciated.

Erik

[lion]

Separate bias, and cut-down gain-up, can be consolidated.

The reason for the odd cut-down/gain-up came about in my attempts to fight the physics it appears. To my initial surprise I can't get more than 3.3V  swing out of U1A without clipping. Some read up seems to indicate that's just how it is with TL072's.

With T1 gain aprox 11 and U1A aprox 1.6 - signals higher than 190mV input will clip U1A (hmmm - is it the opamp input or the output that hit the limits?). I'm not sure if that's a problem here, but I might consider raising the headroom somewhat if possible (without lowering the gains) to be safe. I wonder if another opamp version might help - looking at the datasheets for NE5232, LM833 doesn't tell me much at first glance. Maybe a RRIO. Any suggestions?

Erik

[PRR]

> T1 gain aprox 11 ...- signals higher than 190mV input

Seems likely to me that level will clip T1, before it gets to attenuation and op-amp.

And generally you should not put guitar through gain of 10+ with only 9V supply. Gain of 3 may be all you can stand.

FET with only 9V, and opamps on +/-9V(?), it should be easy for the opamps to pass anything the FETs can deliver.

TL07x amps will swing within 2V of their rails. At +/-9V, 18V total, that is 14V peak to peak, 5V rms. Going to an "ideal" R2R device would give 18V p-p, 6.4V rms. If the one is a problem, the other is not a total cure.

[lion]

Seem like I might have a couple of issues needs looking at - and certainly one around U1A, which really has me confused.

Right now I have one foot out of the door for a 8-10 day holiday. When I get back I'll dive into the details again and try to find where I have gone wrong. Thanx.

Erik

[anotherjim]

When you get back from holiday...

Paul is right, T1 would be clipping and since you had padded down the input to U2a, the output swing of that would appear limited.

Where both sides of a circuit are ground referenced, you do not need DC blocking capacitors. That is all cases between op-amps. The transistor stages are whatever the Drain voltage turns out to be and that isn't going to be ground, so they would need blocking caps.

Size of blocking caps isn't too critical as long as they are not too small. As a minimum, I like to remember this - 100k load needs 100nF. This is a reciprocal relationship, so 10k needs 1uF, 1M needs 10nF. Some will standardise all to the largest instance, it makes purchasing easier.


Pot values depend on both source and load impedance. Op-amp output I'd reach for 10k every time. Transistor stages at least x5, preferably x10 the drain/collector resistor. The load on the pot wiper won't stop it working, but it would be nice if this was also x5 the pot value or the taper of the pot is changed by the load - which tells you what input resistors you need on the following stage. You can overload the wiper to set the range of the pot, which it looks like you've done with the 47k clean pot loaded by 15k.


I'd consider placing the feedback control in the pre-amp. As you have a non-linear amp (NLA) for record, would it not be better for tape like sound if feedback re-records via that same amp? That way, if feedback is set too high, you should get the same build up until it hits the wall that a tape delay does.
If delay time is going to be at least a bit longer than fast slapback, the polarity of the feedback isn't really important (what's 180deg phase shift against 80ms?), so you could use the inverting input of U1a to mix in the feedback.
...otherwise, I can't see what U2a achieves that U1a couldn't also be doing - driving the clean mix clean source as well as the record send. - and U2 would be redundant in that case and U1a driving clean and NLA input.

[lion]

I've been going over the circuit details again, and here's what I've found.

You sure want a gate resistor on T1.

Turned out I had a 1M gate res already. Just missed it in the drawing.
I'll postpone any further revisions/clean up until I've finished researching the problem I seem to have around U1A.

> T1 gain aprox 11 ...- signals higher than 190mV input
Seems likely to me that level will clip T1, before it gets to attenuation and op-amp.
And generally you should not put guitar through gain of 10+ with only 9V supply. Gain of 3 may be all you can stand.

Yes, T1 is doing "something" (which is the reason for having it there), but it's gentle. T1 is basicly the Fetzer, with drain/source resistors set according to the measured Vp/Idds of the particuar J201 in place (ROG recommendations), and I've chosen the Fetzer circuit because I like what is does.

I percieve the guitar sound through the preamps (echo off), as just havings some added "warmth & enhancement" - I take it, due to any added second/even order harmonics from T1 - but still "clean" and natural sounding.

When T1 clips it does it asymmetrical - but, when/how much, obviously depends on input level. Which brings the question: what is my signal level?

I play Strats with vintage type PU's and I have the height of the PU's quite low to get my tone - leaving the output level on the (very) low side. I find it difficult to measure guitar signal amplitude, but I've found that, when using a 1000 Hz sine wave, 75 millivolt seems a good representation of my "average" guitar signal (that is, at this level the amount of sine curve clipping/bending/etc I can see on my scope from a given circuit, seems to correspond to what I'm actually hearing from the circuit - NOT JUST on the signal peaks, but on the "main" part of the sound). Obviously the peaks are higher amplitude, and for that I use a 200-250mV sine wave.

I think all's good around T1. Here's a couple of plots showing the output of T1 at different input levels - and the resultant output levels:







Moving on to U1A.

FET with only 9V, and opamps on +/-9V(?), it should be easy for the opamps to pass anything the FETs can deliver.

TL07x amps will swing within 2V of their rails. At +/-9V, 18V total, that is 14V peak to peak, 5V rms. Going to an "ideal" R2R device would give 18V p-p, 6.4V rms. If the one is a problem, the other is not a total cure.

If I understand this correctly, it seems I have a problem around U1A.

With just below 200mV into the preamp (which gives 1.74V out of T1) the output of U1A starts clipping the sine wave bottoms. Unless there's something faulty with my methode/measuring , that's a max swing out of U1A of just 3.5 volts.



At 300mV input the wave tops hits the limits too:


I've looked at the DC supply to U1. I'm getting the pos. voltage from a regulated supply, and derive the negative voltage through a ICL7660S converter circuit. There's a small conversion loss (which I've read is normal) leaving me with +8.98 and -8.42 volts to be exact. After a a set of 100R/10uF's ahead of the opamps, the voltage at the pins read:
Pin 4 = -7.27V
Pin 8 = +7.85V

I'm quite satisfied with what the complete circuit does as it is - and maybe any clipping in U1A is part of how it's sounds (?) - but I'd like to understand if what I'm seeing from U1A is normal/correct?

To complete the picture, if anyone is interested - here's two plots showing the output to the digital delay with the simulated "tape/head saturation" due to the clipping by the diode ladder followed by T2 (and whatever U1A is doing):



Erik

[lion]

Where both sides of a circuit are ground referenced, you do not need DC blocking capacitors. That is all cases between op-amps. The transistor stages are whatever the Drain voltage turns out to be and that isn't going to be ground, so they would need blocking caps.

Size of blocking caps isn't too critical as long as they are not too small. As a minimum, I like to remember this - 100k load needs 100nF. This is a reciprocal relationship, so 10k needs 1uF, 1M needs 10nF. Some will standardise all to the largest instance, it makes purchasing easier.

Pot values depend on both source and load impedance. Op-amp output I'd reach for 10k every time. Transistor stages at least x5, preferably x10 the drain/collector resistor. The load on the pot wiper won't stop it working, but it would be nice if this was also x5 the pot value or the taper of the pot is changed by the load - which tells you what input resistors you need on the following stage. You can overload the wiper to set the range of the pot, which it looks like you've done with the 47k clean pot loaded by 15k.

Thanks for the guidelines Jim. Very usefull - and kind of what I was looking for.
However I'm not sure I fully understand the very last sentence - about "to set the range of the pot". Can you please elaborate?

I'd consider placing the feedback control in the pre-amp. As you have a non-linear amp (NLA) for record, would it not be better for tape like sound if feedback re-records via that same amp? That way, if feedback is set too high, you should get the same build up until it hits the wall that a tape delay does.
If delay time is going to be at least a bit longer than fast slapback, the polarity of the feedback isn't really important (what's 180deg phase shift against 80ms?), so you could use the inverting input of U1a to mix in the feedback.

Valid point about the feedback. Actually I've tried just that in a couple of previous attempts, but for some reason I couldn't get it to works properly (sound authentic) - much to my amazement and frustration. After a lot of work without success I found it better to treat the FB inside the digital delay to get the necessary degeneration. Present project is much simplified version of my previous attempts - but best sounding so far.

...otherwise, I can't see what U2a achieves that U1a couldn't also be doing - driving the clean mix clean source as well as the record send. - and U2 would be redundant in that case and U1a driving clean and NLA input.

Actually U2 was initially added to drive a SVF (Insert point) for some gentle mid EQ of the dry signal, I'm not yet decide whether to go on with the SVF or not.

Erik

[lion]

Update on my "Vintage-Echo-Sound" preamp project - and more questions

A more detailed investigation into my guitar/pickup signal level proved that it's low indeed, and actually even lower than what I previously thought.

Tacking a peak indicator into circuit, and whacking single notes as hard as I can, I set the indicator LED to flash on the peaks. Subsequently, injecting a 1kHz sine wave, I found the set peak point equaled a test signal amplitude of 53mV. The peak level is practically consistent between N-M-B pickups, and also between my 3 Strats with vintage type single coils. Strumming power chords hard, and even engaging the overdrive pedal on my board, the peak level obviously increases, BUT there's no way my signal can ever reach 100mV (just recorded and noted as a design limit).

On the ground of that, and after some thinking (I wasn't really unhappy with the way the circuit worked - but I'm a curious natur), I decided to try to optimize the preamp for a nominal input level of 30-40mV, 55mV peak - AND no signals above 100mV ever happening - and re-organize the gain structure accordingly:

Lowered the gain of T1 from 11 to 6 by removing the source resistor bypass cap. Instead set up U1B to work with a gain of 5 as a pre-gainstage in front of the JFET. Lifts the pickup signal level nicely to work more in T1's "sweet" range. Still no real "distortion" happening, just a bit more pleasant warmth and "harmonic enhancement".

SW1 added for the option of bypassing the JFET in the dry path, giving the choice of a "direct"  or "harmonics" mode.

Permanently included a parametric EQ cut-only filter in dry path for some gentle tailoring (SVF from Westhost fig. 5: http://sound.westhost.com/articles/state-variable.htm). U2A is the necessary buffer for the SVF.

Changed U1A (previously a gain stage) to just a buffer to keep the effect of the diodes isolated from the dry path. R8 limits the max drive range, P1 controls the amount of "tape-saturation" clipping from the diodes and subsequently T2.



All in all I think the changes made a worthwhile improvement.

As a next/last step I hope to be able to also improve S/N ratio a bit before boxing everything up. Brings up a bunch of new questions.

Trying to learn about noise in circuits I'm studying Douglas Self's "Small Signal Audio Design". To keep Johnson noise as low as possible, I understand it's about keeping impedances low. I'm trying to get my head around it, but I'm getting confused by how it seems to conflicts with our general desire for high Z-in and low Z-out stage impedances. I'm probably mixing up things here - and certainly showing my lack of basic knowledge - and I'm hoping for some expert guidance.

Say I wanted to "silence" the preamp circuit as much as possible (may not be worth it in practice when everythings in the end comes out of a wound up guitar valve amp) if for nothings else just as a learning process. Concentrating on the dry path for a start, and going from back to front:

R15, R16, R17:
The present 15k mix/feedback resistor values are "stolen" from a general article about summings amps. In the SSAD-book there's an in depth chapter on the virtual earth summing amp design. It says reducing the value of the resistors by a factor of 3 will reduce Johnson noise almost 6dB. What I'm missing here is what sets the rules/limits for how low we can go?.

P3, P2:
Mr. Self talks about using buffered attenuators to keep resistors low value and keep Johnson noise down. Isn't P2 and P3 incidentally just that - buffered attenuators - and does it mean that the P2/P3 could be say 1k or less?

R4:
Johnson noise from a 1M resistor is 19microV according to SSAD - for a 10k it's around 2microV in comparison. I don't think R4 could be as low as that, but what would be the optimal value of R4 - and isn't there kind of a conflict here between best noise specs and other performance desires (hope I doesn't sound too stupido)?

Finally R1, R2, R3:
Getting the best S/N ratio out of the first gain stage must be essential. R1 set Z-in, and if a guitar is plugged in directly I understand it's generally desired to be in the mega range. In my case I always have a buffer in the signal chain ahead - so R1 doesn't necessarily need to be that high.  Question - how low is it advisable to go and why - and will it make any practical difference.

As I type I realise I've missed resistors R19/R20 round U3B in the SVF filter - setting a gain of 1 in the inverter stage. Again, reducing by a factor of 3 could cut Johnson noise by >5dB's. Is it as simple as that?

Any help, advise and comments will be much appreciated. Thanks.

Erik

[PRR]

> To keep Johnson noise as low as possible, I understand it's about keeping impedances low. ....seems to conflicts with our general desire for high Z-in

No. Load impedance is not hiss impedance.

Your first stage. The guitar hiss is mostly its ~~40K resistance in the mid-high audio range. This is around 3uV hiss in 20KHz. (Uniformly less if your speaker cuts-off at 5KHz as many do.) The input hiss voltage of 12AX7 -or- TL072 is about 2uV. You could do a little better with a lower hiss voltage device. But stuff like '5532 has high hiss -current- and this in the ~~40K guitar induces hiss voltage. 12AX7/TL072 are hard to beat at guitar level.

Added to that is the hiss in your 6K:1.5K gain-set network. KISS, this is 1.5K, about 0.6uV hiss, and quite small compared to guitar self-hiss or base amplifier self-hiss. (You could go 2K:500 here for a reduction in resistor hiss, but swamped by amplifier and guitar hiss.)

T1 self-hiss is hard to guess, except it is the "same" as a TL072's input stage (JFET at moderate current). Since U1B has gain of 5, T1's self-hiss is totally swamped by what comes from U1B.

T1 amplifies all this hiss, so U2A U1A hiss doesn't matter at all. And whatever could be added by Drive pot and 68K is again probably completely swamped by input area hiss.

The main noise source I see is this Delay permanently summed with the output. Few delays are silent. I'd want to have gain control before and after the delay, to run it as hot as possible without distortion, and let only as much delayed signal back in as needed.

[PRR]

The 1.5Meg at the input appears to be a huge hiss voltage. Maybe 17uV. If you play with guitar not plugged in, it will hiss significantly. But that's pointless. If you connect a ~~40K guitar across the 1.5Meg, the hiss voltage -divides- as a voltage divider. 40K/1500K = 0.027, or 0.5uV hiss from 1.5Meg. This is "small" compared to the ~~3uV hiss of the guitar, or the ~~2uV hiss of the first amp. Also the 40K:1,500K divider drops your guitar level, to 0.975 of "unloaded", which is of course insignificant. Going toward 100K input bias resistor, the 100K self-hiss is lower, but the division against the 40K guitar impedance is much less. Also the guitar signal is down to 0.7 of unloaded, a major hit. (And since guitar varies 5K-100K over the frequency band, response is messed-up.) So yes, that input resistor can be way over 500K and the hiss actually gets less. A Meg or three is safe for FET input chips on bar-room circuit boards.

[lion]

Thanks Paul - much appreciated, helps me a lot.

I'll not pretend I fully understand everything you said, but I'm working on it.

Re low(er) level noises getting swamped by the higher noise of the front section.
Hmmm... - I reconned the first stage and the FET are the main noise sources, but I thought any/all noise elements from any/all sources in a circuit would unavoidably contribute to the final noise level, ie EVERYTHING ADDING UP! Isn't that the case (at least in theory)?

The input hiss voltage of 12AX7 -or- TL072 is about 2uV. You could do a little better with a lower hiss voltage device. But stuff like '5532 has high hiss -current- and this in the ~~40K guitar induces hiss voltage. 12AX7/TL072 are hard to beat at guitar level.

Added to that is the hiss in your 6K:1.5K gain-set network. KISS, this is 1.5K, about 0.6uV hiss, and quite small compared to guitar self-hiss or base amplifier self-hiss. (You could go 2K:500 here for a reduction in resistor hiss, but swamped by amplifier and guitar hiss.)

T1 amplifies all this hiss ....

As T1 amplifies self-hiss from first stage, U1B performance is clearly important!
I understand opamp swapping isn't that straight forward (what might look better on paper, might not hold up in actual circuit - depending on a lot of factors).

Just out of curiousity - I've noticed R.G. more than once has mentioned the LM833 as one of his favorites for low noise guitar level circuits. Datasheet looks good at first impression - but again, there's probably more to it?

http://www.ti.com/lit/ds/symlink/lm833-n.pdf

The main noise source I see is this Delay permanently summed with the output. Few delays are silent. I'd want to have gain control before and after the delay, to run it as hot as possible without distortion, and let only as much delayed signal back in as needed.

I've been looking at the noise in the dry and wet seperately - and so far concentrated on the dry path.

As a prelim solution I've used the output control on the delay to control the return level. The plan is to improve on this in a couple of ways. In the delay the output control sit ahead of a driver stage with a gain of 3. Keeping level at max out of the delay, and instead cutting down right ahead of R15 into the summer amp will also reduce any noise from the delay output driver circuitry. There's plenty of signal at hand - I'll estimate a voltage divider of say 120K:5k pot would leave enough delay signal, and some to spare.

Additional idea to be pursued. The voicing of the echoes includes some heavy high end roll off - down in the 2-3kHz range - all handled in the delays DSP. I think it would make it possible to include a LP filter in combo with the 120k:5k divider - to further reduce hiss from the delays analoge output circuit (without disturbing the echo voicing). Not sure if it'll work, but worth/easy to try if needed.

Erik

[PRR]

> LM833 as one of his favorites for low noise guitar level circuits

Guitar level? Or guitar impedance?

He could be right. Spec-sheets are written to confound.

Hiss current spec is 0.7pA/root-Hz. This is 100pA across 20KHz. Taking "40K" as a guitar's resistance where the ear hears hiss best, it looks like 4uV of hiss voltage. (I could have screwed-up a lot of zeros there.) This is significantly more than the hiss voltage of any of the likely candidates.

If circuit resistance is 10K down, this hiss current largely vanishes behind usual hiss voltages.

Ah- divide voltage by current. 4.5nV/0.7pA is 6.7K. A clue (not proof) that the chip may be low hiss for resistances like 3K to 12K, but maybe less-stellar for resistances like 40K.

OTOH. maybe R.G. has observed that typical '833s have higher hFE and thus lower input currents than what was writ in the specs. I dunno.

[lion]

Thanks Paul,

When I have everything tuned up, including the delay control modefications etc, I might give the LM833 a try. Turns out i happen to  have a couple in a previous build.

Erik