Honey Dripper noise

[durian]

I have built a Honey Dripper and while it basically works, I get a soft whistling sound on the output. This occurs even without an input. I'm hoping someone can offer some advice and help me track down the problem.

Using an audio probe and working backwards from the output, I traced the problem to high frequency (audio high, not RF high) noise on the 18V rail. By removing ICs one at a time, I was further able to track down the source of the noise to IC5 pin 4. If I remove IC5 from its socket, I don't hear any noise on the 18V signal. With IC5 socketed, I hear a high pitched noise. I swapped out IC5 with another LM324, but it didn't help. I've also touched up all my solder joints and I did't see any obvious shorts or opens.

I used the audio probe to check the other pins on IC5. I hear the same type of noise, though sometimes at different frequencies or amplitudes on pins 1, 2, 4 and 14. On pins 3 and 6 I hear more of a white noise type of noise. The other pins are quiet.

Here are some voltages:
18V rail -> 16.23V
10V rail -> 9.46V

IC5 socketed:
1  4.75
2  4.75
3  4.40
4  16.21
5  4.55
6  4.75
7  4.41
8  2.19
9  2.60
10  2.60
11  0
12  2.60
13  2.60
14  0.42

I also checked the voltages with IC5 removed from the socket
1  0
2  0
3  7.03
4  16.48
5  4.56
6  6.53
7  6.77
8  0
9  0
10  2.60
11  0
12  2.60
13  0
14  0

Any advice on what might be going on or where I should check next would be greatly appreciated.

Thanks,
mike

[RobA]

How high is high (1kHz, 3kHz, >10kHz)? The LT1054 does inject noise onto the power rails. The spec sheet rates them at a nominal frequency of 25kHz and I've never measured one at less than 30kHz, but the spec sheet lists 15kHz as a minimum. It's possible you have one that's out of spec I suppose. I'd guess that it isn't your issue if the noise is lower that around 10kHz, but it is something to check.

[durian]

Thanks for the advice. What you say makes sense. I'll try swapping the LT1054. TI's LT1054 datasheet says the output ripple is proportional to current load, so if IC5 is drawing current, the ripple will increase when it is present and decrease when it is removed. Once I determined that removing IC6 reduced the noise, I didn't test removing IC1 to see if it also reduced the noise. According to the datasheet, it looks like replacing C19 with a larger value might help too. The +5V rail has a 100uF cap. Perhaps I should swap C19 (18V rail, 10uF) with C20 (5V rail, 100uF).

[RobA]

When I was testing these, I found 10µF to be pretty good as a cap value. I did find that they were a bit more efficient when using a low ESR or low impedance cap (like these http://www.mouser.com/ds/2/293/e-ps-23382.pdf).

One thing you could do to test if this is the noise source is put a small cap, 5-20pF, from pin 2 to pin 7 on the LT1054. This should raise the frequency of the clock. It could actually solve your issue too, if this is the source of the noise. It does make the LT1054 less efficient and provide less current, but you might still be at an OK level. I don't know how much the HoneyDripper draws, more than 20mA of the 1044's I'd guess, but probably not more than 80mA or so that the LT1054's look to be able to provide when they are clocked up. The text on page 10 of the spec sheet and figure 15 talk about it. Although, figure 15 is a bit confusing because they are describing the two different methods for raising the clock and the method of lowering the clock all on the same diagram. Anyway, it would be worth just touching a 20pF cap across the two pins to see what it does.

[durian]

I've done some more experimenting. I swapped the LT1054, added a 15pF cap across pins 2 and 7 (had to use an 0402 surface mount as I didn't have any through-holes that low) and swapped the 18V and 5V tank caps. None of those changes really helped. I looked at the signal with a scope and did see some noise, but it was about 35kHz. At this point, I figured I probably screwed up when I was tracing the source of the noise and started afresh.

On my second go-around, I have traced the noise to pin 7 of IC4. The signal that connects C6, R16, R17 and the output of the op amp has significant noise. This noise is attenuated on the other side of R17 (towards the circuit output), but is still present enough to be heard. I don't hear it on the other side of C6 or R16.

I tried swapping the JRC4558 with a different one, but it didn't help. I couldn't hear anything on either of the op amp's inputs. There still might be something on the op amp power, but if there is, it is pretty faint. The other op amp in the same package is clean.

I did try swapping out the CA3080's (sourced from Small Bear). I've got some NTE996's showing up today or tomorrow and I'll swap those in when they arrive.

[RobA]

Thirty five kilohertz is up there where the LT1054 normally has its clock and it certainly shouldn't be the cause of your problem. So, I think that is probably clear and the problem is somewhere else.

If you are hearing it on pin 7 of IC4 and not pin 1 and this didn't change when you switched the two 3080's, I'd suspect that it was related to the passives in the filter network of that part of the filter. I'd check through the continuity and solder connections of the parts in that area again.

You could try a different type of op amp in that position too. An LM833 or an NE5532 can almost always sub in for 4558 if you have one of those. Even something like a TL072 might work to test the circuit and see what it does there.

Does moving the fuzz control or the rotary switch change the pitch or nature of the whistle?

I haven't built this pedal, so I'm probably not going to be much help at this point. Hopefully someone with some direct experience with it will come in now.

[durian]

Thank you, I really appreciate the help you've provided. I'll continue to detail what I find in case other people have a similar problem and their searches bring them here. Hopefully, there will be a resolution at the end of all this.

Yes, the rotary switch does affect the noise. I'd been testing with it in the 1 position, but now that I've moved it, I see there is noise on pin 1 as well as pin 7, it's just not very noticeable when the switch is in the first position. Here's how I'd characterize the noise.

Pin 7
switch 1  high pitch
switch 2  mid
switch 3  none (or very faint)
switch 4  same as switch 2

Pin 1
switch 1  none (or very faint)
switch 2  low-mid
switch 3  faint mid
switch 4  same as switch 2

An LM833 op amp didn't improve things.

When I was probing this last go-around, I noticed the noise was much more pronounced if I accidentally brushed my hand against the circuit input. Presumably, I'm bringing in noise which is being processed and amplified. Maybe the circuit is working as it should and I'm just not filtering external noise as well as I should be.

And now I'm back to the power supply.

I'm not sure I had my scope AC coupled last night. Now I do and I see 7kHz (and 14kHz - the scope has some basic FFT capabilities) on the 9V input. I also see this on the 18V supply, but not so much on the 10V. I'll hook this up to a bench supply tomorrow and see if that makes a difference. If I understand the LT1054 configuration, it isn't really regulating the input voltage, it is just doubling it. I don't see any low-pass filtering on the 9V input either. Will any noise on the supply voltage get passed through to the 18V supply? I should also point out that I still have C19 and C20 swapped. On my board, the 18V supply has the 100uF cap while the 5V supply has the 10uF cap. If the bench supply helps, I'm thinking I can get by with a very low value resistor on the 9V input in conjunction with the existing 100uF cap, C17,  and still filter out the 7kHz component from my wall wart.

[RobA]

You are right about the LT1054. It doesn't regulate or filter the power. The only filtering you are getting is what the input caps provide. You are probably best off with the 100µF in the C17 position. One thing that might help in filtering some of the high frequency noise is to parallel a 0.1µF ceramic cap across the 100µF. But, from your other observations, I doubt that it's going to help with the whistle.

Is that 7kHz peak on your bare power supply, or does it just show up when plugged into the pedal?

My assumption based on what the effect type is is that the filters are resonant band pass types to set up the formants. If you have much input noise, they are really going to focus that noise at the frequencies of the resonant peaks. Your observations of the noise characteristics are consistent with that. I don't know what the Q of the filters are, but formant filters tend to be pretty peaked to get a good vowel sound.

Have you tested this in the box or has everything been out of the box? It might actually be a bit better in the box. Other things that might help are all the stuff you would do to tame noise out of a high gain pedal.

[durian]

The bench supply didn't help. It's a linear, non-switching regulator and with a clean output, so while there might be noise coming from my 9V wall wart, I don't think it is the root cause. If that was the main issue, using the bench supply should have fixed the problem.

Probing from the input side of things, I found a lot of noise on pin 8 of IC1. In fact, for a little while I could hear voices in the noise. So either, I need to double-check what I ate for lunch, or it's picking up AM radio.

I'm beginning to think that this is just a noisy circuit. Maybe the noise I'm hearing is to be expected. I did notice that when the rotary switch is in position 1, contact B is left floating and I get a lot of noise going into pin 3 of IC5A. Shouldn't pin 5 of the rotary switch be connected somewhere? It is an input to IC5B, after all.

You asked if the board is in an enclosure. Yes, sort of, not really. It is wired into an enclosure, but I have the back off and the wiring on the board allows it to hinge up so I can get at both sides. So, effectively, it's not really in the enclosure.

[RobA]

It always cracks me up when I get radio signals on stuff. It usually happens when breadboarding high gain circuits and there's no shielding anywhere. Those issues really are usually better in the box. It can help to have a small input resistor (~1k) too. If you can't get the noise reduced other ways, you may want to try that by putting one at the input.

It's a bit weird that you are getting lots of noise when in position 1. That particular lead is floating, but the inputs into the op amp aren't. That lead is just going to a bias voltage when it's connected. Do you have the switch board mounted or are you running leads to it off board? If off board, then it generally helps to have the leads be as short as possible. This goes for the input and output wires as well. In some cases, with high gain pedals and noise problems, going to shielded hookup wire can really help.

You may want to try just popping the back on too and see how much noise reduction you get.

[durian]

I had high hopes that buttoning up the board would sweep the problem under the rug, but it didn't help. Alas. The rotary and pots are soldered to the board. I tried grounding the input on the board and the noise is still present, so I don't think I 1k on the input will help. For the same reason, I'm not sure shielded hookup wire will help either.

If I step back a bit and look at the big picture, I get a high pitched noise, almost a sinusoid, on the output even when the input is grounded. The noise character changes with the rotary switch, being most noticeable in position 1. Position 3 is the least noticeable. The noise does not change with the sweep pot. When I attach a guitar and play a little, the noise changes with the envelope. If I play a note and then stop, the noise starts out low and then slowly rises in pitch.

Smaller picture, if I use the audio probe and trace from the input, the first place I hear noise is on pin 8 of IC1. It is a wideband noise. There is also some noise on IC4, pins 1 (low pitch, faint) and 7 (high pitch, louder). On IC5, I have noise on three of the outputs, 1, 7 and 14, not 8 though. Pins 1 and 14 sound the same, a high pitch, sinusoid type noise with pin 1 being louder. Pin 7 is a middle freq sinusoid type noise. I think the noise on IC4 is coming from the noise on IC5. I don't know where the IC5 noise is originating.

I would consider the level of noise on the output unacceptable, and aside from one other person that was able to fix their problem by replacing their CA3080's, I haven't seen any complaints about this board, so I'm pretty sure the problem is with my build, not the design. I must have a bad solder, a bad component or I miss-stuffed a component.

As per the build instructions, I do have a few socketed discretes. I haven't detected any noise on them, but maybe it'll help if I remove the sockets. I'll also swap back C19 and C20.

[RobA]

You sound like you are on the right track for trying to locate the source of the noise. If you can put up some good photos of the front and back of the board, that might be helpful.

[durian]

Here's a link to photos of the top and bottom. Note, the tank caps are still reversed and I haven't removed the sockets yet.

https://www.icloud.com/photostream/#A3JtdOXmJt1qxk

[RobA]

Thanks for the photos. I haven't seen anything that stands out as a likely cause yet. Have you tried moving down the input gain by lowering R5? Since it's the first stage into the whole filter network and it has an influence on the sweep, it might have a pretty big influence on the total amount of noise. The other one would be R21. It says that R21 increases the intensity of the filter. Which I guess means it raises the resonance of the filter peaks(?). The gain of a highly resonant peak can be really high even if it is only for a very narrow band. So, this could really be amplifying any noise generated at the input of the filters. If you haven't tried raising that resistor value, you could try that to see how much it brings down the noise. Of course, you do want the resonance to be high enough to get a pronounced vowel.

[durian]

I did have R5 at 330k, a change recommended in the build notes, but after trying the 180k from the schematic, I did not see an improvement. I've got R21 at the recommended 2k.

I keep coming back to the power supply. If I look at both ends of R42, one side is the 18V supply and the other is the 10V supply. The 10V side is clean. The 18V side is noisy. This is when I'm powering the circuit with a 9V battery, so the noise is not coming from the external power. In fact, noise is being coupled on to the 9V terminal of the battery. The noise is not present when I check the battery on its own. Only when it is attached to the circuit.

When I use the FFT function of the scope, I see a number of harmonics that vary according to the rotary knob setting, but there are two large peaks that are present all the time. One peak is at about 5.2k and the other is around 8.3k - give or take.

I might experiment later with sourcing 18V from my bench supply directly to 18V rail and see if that makes a difference.