Sound clips of resistor types

[cdemike]

There's a lot of pretty heated argument about whether different component types make a difference in an amp's sound. There are a lot of sound clips comparing capacitors available online, but I wasn't able to find any comparing resistors, so I decided to record a comparison. I decided to use the plate resistor for the first gain stage in a plexi-style build since it seemed like that position would be most likely to show differences, if any, since it's early in the signal chain, and that resistor has a fairly large voltage drop across it (i.e., providing the best conditions to bring out any non-linear behavior for "resistor mojo"). These were recorded over a 15 minute span to minimize differences in the wall mains voltage, degrees of wear in the tubes, guitar string life, control settings on either the guitar or amp, and microphone placement. The setup was meant to facilitate easy changes to further minimize those variables' impact: all 8 resistors were soldered side-by-side on a turret strip, so there were no significant differences in lead dress. B+ fed all resistors in parallel, and I used alligator clips to connect the plate to the other side of each resistor. The resistors were selected from batches in order to be close in resistance. There was still variance, but the tolerances were well within the typical +/- 5% rating most amp builders use with highest value resistor measuring 101.3k, whereas the lowest measured 98.4k (i.e., with the highest resistance reading 2.9% larger than the lowest resistance).

Another goal was to test whether or not wattage rating or manufacturer had any audible impact on the sound. To that end, this test involved several resistors of each kind.

There are several important limitations to this "study:" this was still a very informal test with essentially no objective data (no scope tracings, no noise measurements, no frequency analysis, and no use of a looper to isolate individual takes as a variable). Additionally, prioritization of speed of switching in order to minimize the impact of the variables enumerated above involved trade-offs that may have impacted the sound, namely placement of each on turret strips may have involved a very minor degree of lead dress differences since each resistor was positioned differently along the same turret strip. This was the primary reason why I did not do any noise testing, since minimizing lead dress as a variable would have been extremely difficult without a scope and very precise means of controlling lead length and resistor placement.

Sound clips here: https://soundcloud.com/mike-913267560/r ... comparison

The list of resistors is below. Forgive the irritating formatting, but the hard returns are there in case anyone cares to listen "blinded," prior to viewing the list of resistors.

























-1W Vishay metal film; nominally 100k (measures 98.4k)
-0.5W Vishay metal film; nominally 100k (measures 99.7k)
-1W Vishay carbon film; nominally 100w (measures 98.6K)
-0.5W Vishay carbon film; nominally 100k (measures 98.4k)
-1W generic Chinese ("LTVY") carbon film; nominally 100k (measures 99.8k)
-1W carbon comp (MFR unknown); nominally 100k (measures 100.9k)
-0.5W Kamaya Ohm carbon comp; nominally 100k (measures 101.3k)
-0.5W Piher carbon film; nominally 100k (measures 99.9k)

The first and third riffs were recorded in the order listed above, whereas the second was recorded in reverse order.

[martin manning]

Ruh-Roh.

But seriously, a lot of effort there Mike. All moderately to highly distorted tones, and I wonder about not using a looper to get player repeatability and fatigue out of the results, but you sounded pretty consistent.

[Stevem]

Thanks for taking the time and effort to do this !

I would be super interested in any differences that might take place for instance when a 1/2 watt resistor is pushed temperature wise.

[xtian]

Yes, a looper, or prerecorded guitar clip is mandatory for A/Bing this stuff. Also, first gain stage won't be distorting, so even less color from the plate resistor is likely.

[cdemike]

Thanks so much for the feedback! Totally agree regarding the looper -- it's on the wish list, and I'll get some new clips recorded once I have one to make for a better-controlled test. One thing I thought was interesting in this comparatively uncontrolled test was how similarly they sound. In the course of recording this I thought I could maybe feel some differences in terms of how the amp was responding, but it would be extremely subtle (i.e., possibly placebo). A more scientific test is definitely needed to clarify if that was real or a mirage. To xtian's point, a future test will likely also involve substitutions at the phase inverter in order to maximize signal swing magnitude across the resistor. Maybe the best way would be to have 3 test cases: plate resistor substitutions at the input gain stage, one one of the phase inverter plates, and with substitutions at both. One challenge would be the likely introduction of more variance in measured resistance, and I'm not sure there's a way to mitigate that effect apart from possibly creating a second set of matched resistors to be candidates for PI substitutions centered around a slightly difference measured resistance rather than prioritizing both close measurements in addition to proximity to the nominal 100k value.

Pushing the temperature would definitely be interesting. Would there be a good reason to believe that the specific way resistors get hot would matter in this test? The scenario that comes to mind is whether any non-linearity associated with heat would manifest differently if the resistors were pushed with respect to their power ratings versus simply heated, for example, in an oven.

[Stevem]

More so to simulate the the temp rise that might take place as if the amp was in use and dissipating heat .

The use of a old time electric heating pad which could be folded over the resistors being tested and then looked at with a thermal gun seems a easy safe test out to me.

[didit]

Hello —

Dropping this in to see if it’s anyone else’s cup of tea

https://en.wikipedia.org/wiki/Lady_tasting_tea

Best ..

[martin manning]

We are still talking about this? I am seriously MIF'd! ;^)

This means that to have confidence that there is an audible difference in the sound of resistor types, Mr. Fisher would require eight blind trials with two types of resistors (A and B), where four of the trials had type A installed and four had type B. The test subject would need to be 100% successful in identifying the type, and since there is a 1 in 70 (1.4%) chance this could occur by choosing randomly, confidence would be 98.6%. Correct?

[didit]

We are still talking about this? I am seriously MIF'd! ;^)

This means that to have confidence that there is an audible difference in the sound of resistor types, Mr. Fisher would require eight blind trials with two types of resistors (A and B), where four of the trials had type A installed and four had type B. The test subject would need to be 100% successful in identifying the type, and since there is a 1 in 70 (1.4%) chance this could occur by choosing randomly, confidence would be 98.6%. Correct?

!!

[R.G.]

Good to see some thoughtful testing going on. Good'on'ye.

This thread made me remember the early null testing techniques. If you want to know if two things are different, you can set them up on a level "floor" and measure only the difference. In this case, the level floor might be in the form of a dual triode, both sides of which are fed the same signal. If the two plate resistors are as identical as you can make them, then there ought to be literally zero signal >difference< between the two plate voltages.

As a practical matter, two triode halves are not .....perfectly..... identical, and neither are two resistors from the same manufacturer and batch, even when selected for identical measured resistance. But still - it ought to be pretty good, as you're looking for how closely you can get identical results. You could do things like swapping the plate leads to see if one side was gainier, or one resistor slightly bigger, that kind of thing. Once this kind of works, you could use a differential amplifier to effectively subtract the two plate voltages, and allow yourself a nulling adjustment to even out any minor triode or resistor values. It ought to be that you could adjust the output of the differential amplifier down to a pretty low level. How deep this "null" is would tell you a good estimate of how identical the two triodes and resistors are. We should expect that this simple a setup could do this pretty well, giving a deep null.

Now's the tricky part. You sub in a different brand, type, batch, vintage, yada, yada, yada resistor for one of them. Go ahead, tweak the nulling pot. The depth of the null, the lowest signal level you can get from the different resistor, is a measure of how the newly subbed in resistor differs from the original two resistors. No auditioning needed to just tell "are they different?"

There will always be thermal noise in the setup, no matter what you do. So the thermal noise contributions of the two resistors set a limit on how identical they can measure. But that's OK - we can't tonal differences below the noise floor. And anyway, knowing what the noise floor null level is lets us know when we can stop tweaking that nulling pot.

It seems like a setup like this should be able to tell pretty quickly whether the resistors are, in fact, different in what they do to the sound.

[maxkracht]

I think the easiest method to see what the difference is would be to use a DAW. Send your already recorded guitar into the amp, record the output of the amp. Flip the phase of one sample and play it with the other. The resulting noise is the difference. Easy to measure and stay fairly consistent.

[martin manning]

I think the easiest method to see what the difference is would be to use a DAW.

That has been tried before, and it was not reliable. The same configuration didn't null itself out. Somewhere there is a post with details.

[maxkracht]

That has been tried before, and it was not reliable. The same configuration didn't null itself out. Somewhere there is a post with details.

Arguably, if you can't make that consistent, I don't know what hope there would be for a looper and listening tests. Do you remember if there was a reason why it didn't null? Temperature changes or line fluctuation? I'm assuming the amp is the inconsistency, not the recording setup, unless there was a flaw in the recording setup like room noise.

[R.G.]

Yeah, I thought about that. There would be issues with sampling and reconstruction filters, probably phase differences too. In this case, comparing at analog would make more sense, I think.
Looking for tonal coloration might best be done with a broadband pink noise signal or a thermal noise generator to get some signal to compare in all bands.

Doug Self has some good background on nulling/cancellation tests on amplifiers.

[martin manning]

That has been tried before, and it was not reliable. The same configuration didn't null itself out. Somewhere there is a post with details.

Arguably, if you can't make that consistent, I don't know what hope there would be for a looper and listening tests. Do you remember if there was a reason why it didn't null? Temperature changes or line fluctuation? I'm assuming the amp is the inconsistency, not the recording setup, unless there was a flaw in the recording setup like room noise.

Found it. I was caps, not resistors though: https://ampgarage.com/forum/viewtopic.p ... 31#p471931
IIRC there wasn't any explanation given.