The RV-5 Analysis Thread

[Orvillain]

I'm digging into the Boss RV-5.

First up.... input level analysis. I send a -6dB 500hz sine-tone out from my RME UFXII, into my Signal Arts Reamp box. This gives me a way to pad down the signal going into the pedal, to resemble an instrument level signal. Just means I'm not properly spaffing a high level line signal into the pedal.

The pedal is also bypassed - meaning, signal going through it and through the buffers, but not being processed by the DSP; it is a dry tone. Not reverberated.

Now check this out:

Quiet signal:

Medium signal:

Loud signal:

Even louder signal:

Louder Still:

The Loudest:

What does this mean? Well... it means that the louder the signal going into the pedal the more distortion you're going to get at the front end, before the reverb algorithm even comes into it.

I compared this to a DD-8, DD-500, Source Audio Nemesis, and they were all much the same. The UA Golden however, could take some amount of extra level.

Actual voltages measured:
Quiet: 0.052v
Medium: 0.303v
Loud: 0.600v
Even Louder: 0.900v
Louder Still: 1.31v
The Loudest: 2.493v

So as you can see.... quite a lot of harmonics coming out of the pedal when you really really ram it with signal. But even with a quiet signal, you're still getting that 1kHz 1st harmonic being added. With the loud signal, which is a fair bit louder, you're getting other harmonics coming through quite clearly too. Once you start to get into very loud territory, the pedal starts to add all kinds of harmonics - and remember, this has nothing to do with the algorithms yet. This is just the pure circuitry of the pedal.

Here's a comparison against the UA Golden Reverb, with 2.493v going into it:

This clearly shows the UA Golden can take a lot hotter signal.

But perhaps this distortion characteristic is one of the things we actually like about the RV-5??


For context, pickup voltage outputs vary quite a bit. According to a bit of ChatGPT research:

1. Passive Single Coils (Vintage Style)
RMS: 0.05 V – 0.15 V
Peak: 0.20 V – 0.30 V

2. Passive Vintage Humbuckers
RMS: 0.10 V – 0.25 V
Peak: 0.30 V – 0.50 V

3. Passive High-Output Humbuckers
RMS: 0.20 V – 0.40 V
Peak: 0.60 V – 1.00 V

4. Active Humbuckers
RMS: 0.50 V – 3.00 V
Peak: 1.00 V – 4.50 V

Now if those figures hold true, you could say that in many contexts, your guitar could quite easily put enough level into your RV-5 in order to generate these harmonics.

 

[Devin]

Very nice research, kimg

 

[Orvillain]

So now, we plug into the 'B' input only - which sets the pedal up to be wet only. This is a really cool feature of the pedal and you can get some really nice dense textures this way. But I digress.

Now if we feed a single Dirac impulse into the pedal, and measure the distance between the initial impulse and the first sign of life from the pedal, this should effectively be our pre-delay time:

Now I'm seeing a figure of 50 samples. Which at a sample-rate of 48kHz, is about 1.08ms. So effectively instant, with no significant pre-delay at all.

But this seems suspicious. Maybe this is just the early reflections. Maybe the late reflections have their own pre-delay?? Let's zoom out:

You can see here, there is a scattered and increasingly chaotic rhythmic pattern that decays away, but then about 2650samples later, we see a loud double-spike in the left channel, followed by the same kind of repetitive pattern in the right channel. I believe this is where our late reflections kick in. Which means our late reflections have a pre-delay of 2650samples - or about 55ms (considering my sample-rate is 48kHz)

 

[Orvillain]

(I should say, I've been using the 'Modulate' algorithm all this time)

Now if we loop record a bunch of times with the impulse:

See how uniform the clicks are in each channel, and across each take. This strongly hints that there are no modulation characteristics in the early reflections. It is a static network. If modulation were applied to these early reflection taps, you'd see slight timing shifts in the clicks from take to take:

But we really don't.

The stereo behaviour is consistent too. This means the modulation being applied, is in the late reflections.

 

[Orvillain]

So now we feed in a solid 500hz sine tone into the pedal, for 60 seconds. We get this:

As you can see, because of the solid sine going into the pedal, again 100% wet, this reveals the shape of the modulation:

Now we can do a few things.

Over a period of 60 seconds count the number of visual cycles, which is about 175. 175/60=2.91. This would suggest the modulation rate is 2.91hz. Which is quite fast for a modulated reverb.

Let's double check in two ways.

If we take just the left channel and open it up in Sonic Visualiser we can also take a rough measurement:

A cycle length of about 0.36s - equivalent to about 2.7hz.

But if we stay in Reaper and select a cycle we can check the number of samples:

Using the right channel made this easier. I've got 17844 samples.

So the cycle length in seconds: 17844 / 48000 = 0.37175
Convert to milliseconds: 0.37175 * 1000 - 371.75ms
Convert to frequency: 1 / 0.37175 = 2.69hz

So... we can be reasonably confident our modulation rate is within the range of 2.69 to 2.8hz, lets say.

What about modulation depth?
We can use Sonic Visualiser (on the right channel this time) to compute the pYIN pitch trace for the audio:

You can see the pitch is cycling around our initial 500hz. Its lowest point seems to be about 497hz, and its highest point is about 510hz. The general shape is closer to a triangle than it is a sine-wave.

The effective bipolar modulation depth is about +/- 25cents. Anyone who has ever programmed a supersaw patch on a subtractive synth will know the fattening and lushification effects of these kinds of detunes!!

And just to add, the phase of the LFO more or less looks the same across each channel. There is no offset. But the right channel does seem to be quieter. This is definitely something to do with the algorithm, and not my cables or setup. Because the dry tone does not exhibit any volume differences.

Here is the same sine tone through the regular 'Hall' algorithm:

A totally different pattern.

 

[Orvillain]

At this point I should say, my interest is looking into the Modulate algorithm primarily. With my favourite settings. Those are:

Here is a Spectrogram of a burst of pink noise going through the pedal - left channel only:

And here is one of the single impulse going through it - again the left channel:

Let's take the impulse one and clean up the noise floor a little bit using the thresholding and gain controls for the spectrogram. Let's also make the window larger, and map the data against the mel scale - so a Mel Spectrogram:

What do we see?

There's a clear mains frequency noise floor artifact that we can mostly ignore. Everything below 60hz.

We have a full spectrum wash at the onset. A powerful shotgun blast of frequencies. There is immediate excitation of the entire spectrum. This means the early reflections and diffusion network are broadband, with no pre-EQ control before the tank. In other words - your full guitar signal will go into the reverb; potentially with the additional saturation from the circuitry that we discussed before.

The brightest concentration is in the 1-4kHz range. That's right in the 'presence' range for guitar, which strongly backs up my anecdotal observations about the RV-5 cutting through nicely, and being able to sound present and with clarity.

There is some top-end roll off. 22kHz down to about 7kHz rolls off very quickly, essentially gone within the first 300ms or so.

But there is also some dampening of the low-frequency content. Because we aren't just seeing the top frequencies roll off, but we're seeing the bottom frequencies rolling off too - more or less to the equivalent rate.

At the same time, those mid frequencies always remain louder and more present (read: more colourful, and pushed towards red/yellow when the other frequencies get pushed towards green) - both of these observations point to a band-pass filtering approach within the algorithm.

RE: Diffusion. There's no obvious comb filter stripes or resonant lines in the tail. Just a smeared noise-like cloud. Very high diffusion properties. This indicates the presence of many all-pass filters.

 

[Orvillain]

Here is what I get when I take a 20hz-20kHz sine-sweep through the pedal, and convolve the IR to get an impulse response:

Again, we've got the broadband early burst, the band-pass decay profile, a smooth energy falloff, and no obvious comb teeth, modal buildups, or resonant stripes; again signs of lots of diffusion, equivalent to many all-pass filters.

Those descending curves in the middle of the file, towards the end of the decay. That's some kind of frequency ringing, or residual resonances from the feedback network. I don't believe it is aliasing, because it isn't a reflection of another signal.

But if we boost and rotate the colours slightly:

Top right corner. There is indeed some low level aliasing happening.

Now... the above was done with the sweep achieving a 0.7v peak. But what do we get if we actually do go higher, and we introduce all of that saturation from before?

This is with a peak of 1.6v:

You can see that high frequency aliasing really coming through in that top right corner now.

If I'm reading this right, it implies that the anti-aliasing filters are either not present, not strong enough, or their frequencies are set too high. But also... maybe this is one of the things we like about the RV-5??

If we open the sweep file itself, rather than the resulting IR. We can clearly see the aliasing reflections:

Softer sweep:

More aggressive higher voltage sweep:

Why does aliasing matter?

Aliasing is enharmonic "junk" - except in this case, we already know that we love the sound of the RV-5. The Modulate mode is commonly referred to as one of the best reverbs out there, and I'm not the only one saying this. So maybe that enharmonic junk is part of the love?

 

[Orvillain]

Take away lessons:

- The pedal will distort quite easily. High output humbuckers into a RV-5? You're going to get some extra saturation from the pedal alone. It won't be "audible" as such, but it creates frequencies that your amplifier will see and process.
- The modulation is not subtle. It is a triangle-ish LFO around 2.7 to 2.8hz. The depth is fairly wide; +/- 25cents.
- The tank is broadband at onset. The full signal slaims into the reverb with no pre-EQ filtering.
- There is a band-pass voicing circuit inside the tank. Lows are damped quickly. Highs above 8kHz vanish within 300ms. The 1kHz to 4kHz band dominates.
- The algorithm is very mid-focused because of this.
- The decay tail is a dense smear, not comb-like.
- No obvious resonant stripes, meaning heavy use of all-pass filters. This equates to a very high diffusion.
- This is why the Modulate mode sounds smooth and lush, but also synthetic compared to a real room.
- Sweeps clearly show aliasing artifacts.
- The aliasing folds back into the audible range, contributing to grit. Part of the overall character.
- There is no major phase offset between the left and right sides.

The RV-5 is not high fidelity. In fact, it is quite low quality compared to reverbs of today. It came from the early 2000's DSP-era where fixed-point math was still largely in use, and where the pressures of market forces coupled with limited DSP techniques, it has resulted in quite a nasty and characterful sounding reverb.

 

[Orvillain]

Let's compare against Helix's Dynamic Hall.

Here's the single impulse going through Dynamic Hall. The top one is with the low cut and high cut disabled. The bottom one is with the low cut set to 1kHz. Both of them have dampening set to 10kHz, and diffusion at 70%: Motion is completely turned off at 0%.

The first thing you notice is, the low-cut is quite subtle. Even during the tail of the file, there is still some energy left in the lows. Also, the low-cut (and presumably the high-cut) occur after the reverb processing; not inside the reverb tank. This means they're not even truly part of the reverb algorithm, but rather a post processing step. You can tell this, because a 1kHz low-cut inside a reverb would result in something even more drastic than the RV-5 has above.

Here is what we get when we use the 'Low Gain' parameters. The gain is at -2dB, and the frequency is at 500hz.

You can see that as the sound progresses, more and more low-end is cut out of the full body of the reverb tail. This implies that the low gain is a low frequency dampening filter of sorts; probably a shelving filter set to 6 or 12dB.

(Note aside: the fact these these two parameters and the main dampening parameter are on different pages on the hardware, actually makes this algorithm much more difficult to dial in.)

Now checkout that against the RV-5:

If you ignore the noise-floor of the RV-5, you can see quite clearly that the dampening filters inside the RV-5 algorithm are much stronger than what you get inside Dynamic Hall.

The bottom is the most extreme the low gain circuit can get. -15dB at 500hz. You can really see how it very quickly filters out low frequencies within the tank.


Now let's go back to just the full Dynamic Hall without the post filtering, and the dampening still at 10k:

And let's talk about this:

What on earth...

First of all, we're seeing a reflection here. From about 12k upwards, as the energy in the reverb is taken away, and high frequencies are attenuated... for some reason above the dampening frequency, they're added back. This screams "BUG!!!" to me, but perhaps not.

These stripes are not musical harmonics. They're byproducts of the algorithm, or the inherent limitations of the HX system. The stripes themselves (the dark lines) remain static in frequency.

Here is the same thing, but versus setting the Dampening frequency to 1kHz:

The nature of the lines and the energy in that region, does change, when you tweak the dampening frequency.

Let's turn off the Low Gain circuit altogether:

That region around 12k and upwards, is quite concerning to me. You can see how with dampening set to 1k, the reverb tail is suppressed very quickly. But we still have energy occurring where it theorhetically shouldn't.

This look like aliasing to me.

So in short - with Dynamic Hall, you do have the ability to dampen high frequencies and low frequencies separately. But it is a bit difficult to control due to using multiple pages on the hardware, and also the low frequencies having two parameters while the high frequencies have one parameter. The names also do not match and unless you did these kinds of tests, you might not even know what they do. There also seems to be a DSP bug or limitation somewhere that results in strange frequency banding above 12k.

 

[Orvillain]

A few quick traces of some other reverbs. I have not tried to dial these to be close to the RV5.

UA Golden Hall 224 A. Bass rolled off a bit. Treble boosted. No modulation. Longest tail possible:

MXR Reverb. Mod mode:

MercuryX. 78 Hall. Bass at 2%. Mids at 80%. Treble at 93.8%:

MercuryX. Cathedra. Lo Freq 0%. Hi Freq 100%.

 

[Orvillain]

Last thing. A four way comparison:

Turns out, the RV-5 'Modulate' mode is quite unique, difficult to reproduce on other bits of kit, and as such, I think I'm stuck with it.

PS: The RV6, RV500, and RV200 do not sound the same as the RV-5!

 

[2112]

Nice work @Orvillain

The "shortcomings = character" aspect of this pedal reminds me of a bunch of old rack effects I've got. They're cumbersome to use (and often very frustrating) but they definitely scratch very specific itches.

 

[Orvillain]

Nice work @Orvillain

The "shortcomings = character" aspect of this pedal reminds me of a bunch of old rack effects I've got. They're cumbersome to use (and often very frustrating) but they definitely scratch very specific itches.

Yeah definitely. I feel like while the older gear has more issues in terms of signal quality, noise floor, and affecting the dry signal too much... they quite often sound more characterful and a bit more musical in some ways. There's definitely room for a nice smooth reverb like what you get from Chorus Hall on the VP4 for example, but the RV-5 is just fiiiiiilllthyyy in a good way.

 

[the swede]

Yes, it’s the ”shortcomings” (compared to whatever modern things…. But why compare) that makes the RV-5 shine. In the same way the shortcomings of the long chip DD-3 also has its following. It’s very romantic in a way that you go to these lengths to find an answer to… “why do I love this pedal?”… when it’s no point questioning it.

 

[Jarick]

Sir, you did not disclose your status as an engineer.

 

[Jarick]

How's the MXR Reverb sound? I've been curious about that one a while.

 

[AlbertA]

Nice analysis.

 

[FuzzyAce]

So what's your endgame Orv with all this analysis, you thinking of making a plugin model or something else? I've never tried the RV-5 or any Boss reverbs to be honest. I was close to getting an RV-7 at one time but the HoF was good enough for my needs.

 

[SillyOctpuss]

I love threads like this. Absolutely fascinating and need to be read a few times to really understand them.

 

[Orvillain]

So what's your endgame Orv with all this analysis, you thinking of making a plugin model or something else? I've never tried the RV-5 or any Boss reverbs to be honest. I was close to getting an RV-7 at one time but the HoF was good enough for my needs.

Originally I just wanted to figure out what was going on in the algorithm, as much as I could, so I could easily program a similar thing (and much closer thing than I've done before!) into the Axe FX III.

I haven't really managed it yet. But sending in similar test signals into the reverb block on the Axe FX III does reveal some interesting things.

I might look into doing my own version of it. Essentially based on a 4x4 feedback delay network using a Hadamard matrix. But we shall see.