Resistors and Hiss

[Justin Thomas]

Hey All,

I'm trying to dump some signal in a build, using the voltage division trick like all those uber-gain amps do, using some high-value resistors. I've read that series resistance can add hiss, and that's what I'm getting: LOTS of hiss, not affected by any of the controls and that wasn't there before I added the voltage divider. So I'm still experimenting with ratios, but I'm wondering - how low can I go for the values, in an attempt to cut the added noise? Most values I see used are between 220k-1M. But could I go lower? Like a 100k & 50k? It does seem as though the hiss gets less with lower values; I also notice some amps add a small cap in parallel with the resistor - is that to assist with noise?

This is a fairly simple amp, basically a Fender Normal channel feeding an LTPI & 2x6L6s. It seems really hot (loud!) and I'm trying to tame that a bit. There's not much in the way of places to add the voltage divider, so right now I have it after the Volume control and before the 2nd gain stage. I tried adding it at the PI input, but that did some REALLY don my stuff... I'm thinking my only other choice is right BEFORE the tone controls, immediately after the 1st gain stage; maybe? Yes? No?

I'm open to other ideas, such as changing plate load resistors to a lower value, and other fairly simple tweaks. It's a real PTP build, so it's kinda rough doing radical brain surgery... Thanks!

Justin

[The Dude]

Schematic?

[Justin Thomas]

Dammit, I saw that coming...
Will have to wait, as this thing's been changed so much, there's not much left of the one I have, before the PI...

Justin

[The Dude]

Well, I got nuthin' without seeing what you built. I'll say this: It's about signal to noise ratio, so the later in the signal chain you cut gain, the less noise. You want to keep the signal level as high as practically possible as long as possible so that your not amplifying all the noise.

[Justin Thomas]

Good theory to work with. Thanks! I actually just thought of a problem I may have caused when I had it at the PI, so I will revisit that, also. Say I just put a 470k to ground off the PI input grid; I basically lost all of my signal, and then could only get any if I really whacked the guitar hard. I'm guessing it had something to do with input impedance or whatever, which I admit to having absolutely NO CLUE about... So I'll try a true voltage divider there tomorrow & see what I get.

Part of this is experimentation, and I've got about 30 days to get it right; it was good as was, but I can do better... I am fairly certain he'd like a LITTLE bit more useful range on the Volume knob.

And I'll try & get that schematic to y'all. It'll be hand-drawn, I promise.

Justin

[pdf64]

I think that the key thing with resistor hiss is the impedance between the grid and 0V; so from the perspective of the grid, what is the combined total impedance of all the parallel paths back to 0V.
So via the grid leak / pot wiper will be one.
Via the plate resistor of the preceding stage will likely be another (unless perhaps a pot is turned very low).
Obviously a grid stopper will add to this.

From that, it may be apparent that signal attenuation by means of a split load plate resistor on the preceding stage may often provide a good means of keeping noise low whilst allowing centre bias (of the preceding stage).
See 2nd stage of 6G3 http://schems.com/schematicheaven.ne..._6g3_schem.pdf
The downside of this method is that the ratio of signal : 'whatever crap is on the HT node eg ripple' is worsened.

Another method is to just load down the preceding stage's output with a low value of grid leak, as per the 2nd - 3rd stage of the Express http://schems.com/schematicheaven.ne...ck_express.pdf
Handy if cutting some low end is also on the agenda.
Personally I perceive that the tone, especially clean, suffers when a regular 12AX7 stage is loaded much less than 100k.
Hence the normal channel of non-rev BF amps always sounds better (not just louder) than the trem channel.
And BF Fender reverb channels sound better with a 100k (rather than 50k) trem intensity control.

[Malcolm Irving]

... a problem I may have caused when I had it at the PI, so I will revisit that, also. Say I just put a 470k to ground off the PI input grid; I basically lost all of my signal, and then could only get any if I really whacked the guitar hard. ...

Yes. The input grid of an LTPI is at an elevated DC voltage (due the tail resistor) so you shouldn't put a resistor to ground on it.

Another way of attenuating is to replace a plate load resistor with two resistors in series - you can then take an attenuated signal from the junction point of those two resistors. I've never tried it - but it should give you attenuation without any added noise.

[Dave H]

If the first two stages have large (22u) cathode bypass capacitors you could remove one or both of those for a 6dB or 12dB gain reduction. I prefer to do that rather than generating gain I don't need only to have to dump it with potential dividers.

[Chuck H]

Malcolm, the series resistors you describe are called a split plate load. Pete covered it in his post above and mentioned the drawback of potentially increasing noise from the HV rail getting into the signal.

Dave, WRT bypass caps I wouldn't suggest anything but full bypass for the first gain stage. It's not only a risk of hum, though proper grounding seems to prevent this just fine, but there is also a lot of emission noise on the cathode. A tube doesn't even need to be gassy to do this (I tried several in the following experiment). On a higher gain build than Justin's I found that full, rather than partial bypass on the first gain stage eliminated emission noise, which can sound a lot like Johnson noise. This being my experience I think it's better to keep any bypass cap changes to the second stage.

Justin, I agree with Dude (including the need for a schematic ). You really do want to keep gain up if possible on stages that never reach clipping. That would be the entire preamp in a Fender BF "normal" channel. Any attenuation should happen after the second gain stage. IME the split plate load works fine here and I haven't noticed any rail noise. And why would I? The DC is pretty damned pure by this point in the rail.

Malcolm called it on the signal drop when a resistance is referenced to ground at the PI grid. The LTPPI grid is isolated from DC influence by a coupling cap BEFORE it's bias is established by the grid resistors references to the elevated tail (Hope that gets an "ah ha" moment). So any signal attenuation before the PI would need to be done upstream of the coupling cap.

You really don't need to use high resistance for dividers and loads. YMMV but I haven't had the same trouble as Pete loading 12AX7's below 100k. You probably wouldn't need to go below 100k anyhow. If you used a divider with a TOTAL resistance of 100k then that would be the load. My vote stays with the split plate load at this time though. It get's the job done clean and requires only one additional resistor over a non attenuated arrangement. I typically put a 100k linear pot in place as the plate load and take the signal from the wiper, then adjust the pot until the amp is doing what I want. Then take out the pot and measure (hoping it's close to standard resistance values. Staying close to 100k plate load your options are pretty much 18k/82k 22k/82k 33k/68k 47k/47k (and, of course their inversions). You can get more finite resistances in tighter tolerance parts but I generally try to stay with the most common values in my designs wherever possible.

Looking forward to the schematic.

[Dave H]

Dave, WRT bypass caps I wouldn't suggest anything but full bypass for the first gain stage. It's not only a risk of hum, though proper grounding seems to prevent this just fine, but there is also a lot of emission noise on the cathode.

I was aware of the hum issue but I didn't know about emission noise. Neither seems to be an issue on the low gain amps I make. If an amp distorts before half rotation of a 10% log vol pot I call it high gain.

[bob p]

Dave, WRT bypass caps I wouldn't suggest anything but full bypass for the first gain stage. It's not only a risk of hum, though proper grounding seems to prevent this just fine, but there is also a lot of emission noise on the cathode. A tube doesn't even need to be gassy to do this (I tried several in the following experiment). On a higher gain build than Justin's I found that full, rather than partial bypass on the first gain stage eliminated emission noise, which can sound a lot like Johnson noise. This being my experience I think it's better to keep any bypass cap changes to the second stage.

This. I'm also in the camp that the first stage should be fully bypassed in order to get the highest possible S/N with the lowest possible distortion. To that end, if you're planning on bypassing with caps and you want to avoid that wooly sort of bass distortion, big caps are a good recipe ... or better yet LED cathode biasing.

Something to consider if you have an unused half triode in your design is to use a parallel first stage to maximize clean S/N, and then throw away gain later on if need be. IMO there's no substitute for maximizing S/N early on.

[bob p]

You really don't need to use high resistance for dividers and loads. YMMV but I haven't had the same trouble as Pete loading 12AX7's below 100k.

I guess there are a couple of ways to look at this. One is that if you generate too much gain, just throw it away with dividers. And deal with hiss. Another option if you're generating too much gain is to redesign your stages so you just don't generate so much gain that you have to throw some away.

If Justin is in the situation where he already has more gain than he needs (and he's trying to throw away gain) then it makes sense to consider a change in plate loading so that less gain is produced. I've used Ra=47k with a 12AX7 with good results. In doing that, the output voltage swing is going to decrease quite a bit (check the load line) but then current output will increase. Decreasing the plate load from 100k to 47k will also increase second harmonic distortion by as much as 50% -- which might be what Pete was getting at. All things considered, I think that the lower plate load of around 47k makes the stage sound warmer and more compressive, which may or may not be what Justin is after.

Another option is obvious -- tube rolling. If a 12AX7 generates too much gain, try a lower mu triode.

When it comes to resistor noise, IMO that's a situation where it pays not to use the cheap house brand resistors, and instead read the data sheets to select the best name-brand low-noise metal film resistors that you can get. As it turns out, not all resistors are created equal, and as you go up in value there can be huge differences between generic house brands and the name-branded low-noise resistors. IMO that makes it worth paying more for the quiet resistors in high values. My experience has been that the generic Chinese resistors (house brand for Newark/Farnell as "Multicomp") get really bad at high resistance values and should be avoided. They don't even list noise specs on their spec sheets, which tells you something...

[Dave H]

This. I'm also in the camp that the first stage should be fully bypassed in order to get the highest possible S/N with the lowest possible distortion.

Since when did we want the lowest possible distortion in a guitar amp?

I like the parallel input stage. I have it on my 18 Watt. I usually run the cathode resistor not bypassed but have the option to use a footswitch to bypass it with 22u for a boost.

What do we have to reduce gain? Potential divider, split load, lower mu tube, un-bypassed cathode resistor, plate loading etc. How about using some local negative feedback?

[pdf64]

I think that any signal / noise at the cathode will add to that present at the grid. That would include noise generated by the cathode resistor as well as that generated by the cathode itself. The series resistance value at the cathode will tend to be much lower that that at the grid, but the current will be massively higher, so it may be that the potential for an unbypassed cathode to introduce noise is significant.

Anecdotally, whilst tinkering around with these things, I seem to remember noticing on several occasions that bypassing an input stage cathode noticeably reduces hiss, and with some tubes, reduces hum too.

[mozz]

Going by the old resistance coupled amplifier charts, reducing the plate voltage will also reduce some gain, not much but worth a try?