Topic: Emulating vacuum tube buffer in DSP?

This is just for fun thought experiment, I don't expect you guys to actually do it, you probably have more important things to work on, but I'm curious if it's actually possible?
I have home "studio" that I use for work as well as fun, I recently got into Persona games (3&4) I played them on PS2 connected to the RME dac via toslink and the audio from those games was so riddled with compression artifacts that it was painful to listen to on my system. So I did a bit of research and bought tube pre-amp and a couple of relays to bypass it from ebay for 20$, I also bridged volume control on it because I use RME as pre-amp, and did it make a difference! Not only that highly compressed audio was now completely listenable, it also made poorly recorded music (like early Nightwish albums) a lot more enjoyable, don't get me wrong I would never use it on properly recorded stuff, and especially not for mastering because it strips out the details and adds color, but it certainly was an eye opening experience that better equipment isn't always proportional to enjoyment of sound.

2 (edited by KaiS 2020-03-31 07:16:14)

Re: Emulating vacuum tube buffer in DSP?

Tube emulation, although it looks technically easy in the first place, is not a simple task.


There are so many variations of tube powered devices, each with it's own, distinctive charachteristic:


• Look at the tube version of the German's broacasters workhorse tape machine, the Telefunken M10: a chain of a multitude of tube circuits, all do use a lot of NFB to control their behavior.
Result - not the slightest hint of "that" tube sound you would expect. This is, of course, on purpose, the broadcast sound has to be natural, uncolored, transparent.

• The other extreme: your 20$, underpowered single triode chinese buffer amp of probably questionable design:
All kinds of distortions, limited dynamic range and frequency response:
Seems just exactly what's needed for your purpose, but should this be the model for an emulation?


A lot of these emulations of all kind are already available as plugins.
Sucessfully and detailed emulations of all time classic studio devices —
UAD's LA2A comes into my mind.
UAD builds both, the original hardware still, and their own emulation that uses quite some DSP power.


One may not forget, a tube circuit more then a solid state one, largely depends on it's surroundings, levels etc.

A traditional tube power amplifier interacts, not controls a loudspeaker or headphone.

The sound of the VF14 tube inside the grand-daddy of all tube powered microphones, the famous Neumann U47, is unrivaled even by it's nearest achestor, the EF14 tube.
The reason again is an interaction, with a special power scheme - and the output transformer used.


I went the other way:
Designed and built special purpose tube amplifiers.
E.g. a 1W OTL SE triode headphone amp for the AKG K1000, and a comparable design, a 10W version to drive my Electro Voice Sentry III up to earsplitting levels (what I never, OK, almost never do).

Very low NFB gives a pure triode characteristics, tube as pure as tube can be:

Lot's of harmonics, 2nd order spiced with a bit of 3rd oder only, softly develop over a exceptional wide level range, enriching the sound in a "natural" way.
Coupled with the super precision and cleanliness in the lower level range, typical for OTL- (Output-Transformer-Less-) tubeamps.

These amps have a coincidence with the headphones / speakers driven, not by chance but by design.

Your "tube buffer" likely has such a coinidence too.


My idea is: if you want the real thing you need the real thing, at least it's the shortest shortcut to this destination - and it's fun!

Re: Emulating vacuum tube buffer in DSP?

KaiS wrote:

• The other extreme: your 20$, underpowered single triode chinese buffer amp of probably questionable design:
All kinds of distortions, limited dynamic range and frequency response:
Seems just exactly what's needed for your purpose, but should this be the model for an emulation?

Well that is exactly what I want; There are a lot of young female characters in persona 4 and every time they speak my ears bleed, especially "sss" sounds fall apart with that low bitrate ringing, it's not like It disappears when I switch to tube buffer, you can still hear the ringing but it sounds like it's far away and a lot less ear-splitting; on a first few Nightwish albums all of the instruments sound like they are on top of each other and dynamic range is non existent, but on tubes dynamic range improves a bit and there is some separation between the instruments, also everything sounds a bit more bass-y I guess.
I'm not going to pretend like I understand what my tube buffer does but it makes low quality audio more enjoyable, unfortunately it takes about 10 min of warmup before it starts to sound good and it gets really hot, it would be more convenient and a really cool feature if it was done in DSP but is probably not gonna happen.

4 (edited by KaiS 2020-03-29 06:16:30)

Re: Emulating vacuum tube buffer in DSP?

Have you tried ADI-2's equalizer to do something similar to that tube amp?


Female S-sounds usually cover the range from 5-10kHz.
Try:
Band3 G:-3.5  F:6.3k  Q:2.6
Band4 G:-3.5  F:8.0k  Q:2.6

Or:
Band5 G:-5.0  F:4.5k  Q:1.2  < (shelving)

Or combine both, with G- (Gain-) settings to you liking.

These figures are meant as a starting point for you, largely depend on the headphones used.
Lot's of models have a boost right there that sounds good with well recorded music, but may contribute your problem.


The other part a tube amp does, distortion, cannot be emulated with ADI-2.
But distortion doesn't help with over-exaggerated S-sounds anyway.

Re: Emulating vacuum tube buffer in DSP?

KaiS wrote:

Have you tried ADI-2's equalizer to do something similar to that tube amp?


Female S-sounds usually cover the range from 5-10kHz.
Try:
Band3 G:-3.5  F:6.3k  Q:2.6
Band4 G:-3.5  F:8.0k  Q:2.6

Or:
Band5 G:-5.0  F:4.5k  Q:1.2  < (shelving)

Or combine both, with G- (Gain-) settings to you liking.

These figures are meant as a starting point for you, largely depend on the headphones used.
Lot's of models have a boost right there that sounds good with well recorded music, but may contribute your problem.


The other part a tube amp does, distortion, cannot be emulated with ADI-2.
But distortion doesn't help with over-exaggerated S-sounds anyway.

It's not the "s" sounds that bother me, but the compression artifacts that come along with it. I don't know how to describe low bitrate ringing if you haven't heard it, the only thing I can think of is trying to make CRT display sing by auto-tuing the sounds it makes.
I could also hear it on all of my equipment from passive hi-fi speakers to active monitors, and from Sennheiser hd598 to Focal Elear and Audeze LCD-X.

6 (edited by KaiS 2020-03-29 14:35:21)

Re: Emulating vacuum tube buffer in DSP?

I know that sound, but then it's really low bitrate, like 4bit/8kHz.
Kind like cellphone with very bad connection.
High-pitched tones singing along S-sounds.


Hard to imagine and understand that anything is still produced that way nowadays.
A relative of mine was developing engineer for a big cell phone manufacturer, his task was to squeeze as much sounds into the then limited memory without creating such artifacts.


It's funny that a tube device effects that positive.
Would be hard to find out why, and how this could be emulated.


What if you try a steep low pass filter?
But this wouldn't help if the artifacts are located in the upper mid range, below the S's, like with cell phones.

Maybe looking at ADI-2's analyzer could give you a hint which frequencies to boost and which to cut.
Start with a high-Q 12dB notch and tune it's frequency until the annoying frequencies are best reduced.
Counter act with a broad, slight boost at & above that frequency to regain intelligibility.

7 (edited by Sebastian.Athea 2020-03-30 02:20:04)

Re: Emulating vacuum tube buffer in DSP?

KaiS wrote:

I know that sound, but then it's really low bitrate, like 4bit/8kHz.
Kind like cellphone with very bad connection.
High-pitched tones singing along S-sounds.


Hard to imagine and understand that anything is still produced that way nowadays.
A relative of mine was developing engineer for a big cell phone manufacturer, his task was to squeeze as much sounds into the then limited memory without creating such artifacts.


It's funny that a tube device effects that positive.
Would be hard to find out why, and how this could be emulated.


What if you try a steep low pass filter?
But this wouldn't help if the artifacts are located in the upper mid range, below the S's, like with cell phones.

Maybe looking at ADI-2's analyzer could give you a hint which frequencies to boost and which to cut.
Start with a high-Q 12dB notch and tune it's frequency until the annoying frequencies are best reduced.
Counter act with a broad, slight boost at & above that frequency to regain intelligibility.

I think you're confusing bitrate with sampling rate. PS2 works with 48khz/16bit, however due to DVD media limitations (4.7gb) voices are compressed to low bitrate mp3 or aac or something similar; there are about 30 hours of voices and they have to fit on a DVD with the rest of the game.
Also I don't think it's eq issue, I have tried messing with eq but it didn't help much or it cut the treble, and everything started to sound muffled, but that's not what tube does, you can still hear the ringing on tubes but it's far away and not in your face, while most of the treble is still present.

8 (edited by KaiS 2020-03-30 07:47:25)

Re: Emulating vacuum tube buffer in DSP?

No, I'm not confusing the two.
If you downsample to 8kHz, data compressed audio (variable bitrate) becomes more compact, even if stored or played at 48kHz.

MP3 encoders do this automatically if you set them for lower bitrates.
Nominal (playback) samplerate does stay at 48kHz, but higher frequencies are removed to reduce file size.


If EQ in a way I proposed (narrow band cut that don't muffle the sound much) doesn't help I guess you need to stay with the hardware.

The question that arisis then is, is it really the tube that makes the difference, or something else in this device.
I'm heading for the idea that the tube might be obsolete and could possibly be removed/bypassed, doing away with warmup / heatup.


What's the name/type of this buffer amp, is it still available?
Maybe I can find a schematic.

Re: Emulating vacuum tube buffer in DSP?

KaiS wrote:

No, I'm not confusing the two.
If you downsample to 8kHz, data compressed audio (variable bitrate) becomes more compact, even if stored or played at 48kHz.

MP3 encoders do this automatically if you set them for lower bitrates.
Nominal (playback) samplerate does stay at 48kHz, but higher frequencies are removed to reduce file size.


If EQ in a way I proposed (narrow band cut that don't muffle the sound much) doesn't help I guess you need to stay with the hardware.

The question that arisis then is, is it really the tube that makes the difference, or something else in this device.
I'm heading for the idea that the tube might be obsolete and could possibly be removed/bypassed, doing away with warmup / heatup.


What's the name/type of this buffer amp, is it still available?
Maybe I can find a schematic.

You learn something new every day, I don't work/use with lossy formats so I have no idea how they work.
I think tube is doing something, you can hear sound changing as it's warming up.
I bought my pre amp as a diy kit, just type "tube audio buffer" into ebey

10 (edited by KaiS 2020-03-30 15:08:40)

Re: Emulating vacuum tube buffer in DSP?

I see a bunch of different ones on Ebay.
They all are tube amp only, not tube/solid state hybrid.

So it's the tube doing their thing.


I still don't get how it can suppress or cover the artifacts.
If you are capable of soldering you might have the possibility to measure the frequency response.


All you need is an AC voltmeter (dB range if available) that covers 20Hz-20kHz, which can be tested with direct connection to ADI-2's output, any free tone generator App as source.
If there's a deviation you could even built a correction table to "calibrate" the meter.
The ADI-Analyzers frequency scale scheme is practical.

Then insert that tube amp and take notes of the levels deviations.
Important: use your normal load, e.g. the headphone.
But don't kill your speaker's tweeters if that's the load!
Better not use the speakers.

As the tube can act level dependent use your "standard" listening level, and maybe two more measurements, about 6dB above (double level) and 12dB below (quarter level).


I'd be really curious about the result.
If significant, this might help you to build you own "Emulation".

Re: Emulating vacuum tube buffer in DSP?

KaiS wrote:

I see a bunch of different ones on Ebay.
They all are tube amp only, not tube/solid state hybrid.

So it's the tube doing their thing.


I still don't get how it can suppress or cover the artifacts.
If you are capable of soldering you might have the possibility to measure the frequency response.


All you need is an AC voltmeter (dB range if available) that covers 20Hz-20kHz, which can be tested with direct connection to ADI-2's output, any free tone generator App as source.
If there's a deviation you could even built a correction table to "calibrate" the meter.
The ADI-Analyzers frequency scale scheme is practical.

Then insert that tube amp and take notes of the levels deviations.
Important: use your normal load, e.g. the headphone.
But don't kill your speaker's tweeters if that's the load!
Better not use the speakers.

As the tube can act level dependent use your "standard" listening level, and maybe two more measurements, about 6dB above (double level) and 12dB below (quarter level).


I'd be really curious about the result.
If significant, this might help you to build you own "Emulation".

I know how to solder but I'm not electronic tech, I ask on forums how fix a fault, people tell me what to measure/replace and it usually works.
I have a multimeter but I borrowed it to a friend and due to the current situation I can't leave the town without permit, and it's a pain to get one. But even without measuring it I can tell you that it's colored with multiple peaks and dips although they aren't too high, you can hear them but it's nothing too major. I don't think it has anything to do with the way it sounds though, if I had to guess I would say it has something to do with phase (phase cancellation?), like the effect humbucker pickups have but less pronounced.
If you have the gear you can buy one off the ebay and measure it your self, you'll do a better job than I would.

12 (edited by MetalHeadKeys 2020-03-30 17:54:44)

Re: Emulating vacuum tube buffer in DSP?

Hello, everyone!

My guess is that the tube adds harmonic distortion(the pleasant one smile ) along with a high frequency roll-off. That could help get rid of these artifacts, because the bass and low-mids frequencies become more pronounced! Also, depends on what kind of tube is used, and how hard it is driven!

Korg Triton Extreme from 10 years ago, used a vacuum tube in its design, for adding drive to the sounds, but also for removing artifacts in the high frequencies, caused by aliasing (because every sound was created by samples)!

Re: Emulating vacuum tube buffer in DSP?

MetalHeadKeys wrote:

Hello, everyone!

My guess is that the tube adds harmonic distortion(the pleasant one smile ) along with a high frequency roll-off. That could help get rid of these artifacts, because the bass and low-mids frequencies become more pronounced! Also, depends on what kind of tube is used, and how hard it is driven!

Korg Triton Extreme from 10 years ago, used a vacuum tube in its design, for adding drive to the sounds, but also for removing artifacts in the high frequencies, caused by aliasing (because every sound was created by samples)!

I think tube in my setup isn't driven hard, volume control on tube pre is bridged but on RME my volume is between -33 and -26 db, VU meters peek at -14db

Re: Emulating vacuum tube buffer in DSP?

Hello Sebastian!

It doesn't matter if it is not that hard-driven, because it's allready getting driven by its internal voltage! That's why the effect is taking a bit to become apparent. It's untill the tube gets warm! Thus driven!
I believe that if you raise the volume on the RME, it will become even more apparent!

Re: Emulating vacuum tube buffer in DSP?

I can't raise the volume more because all of my systems are power amps only including the headphone one, I didn't notice much of a difference (if at all) of sound characteristics on lower volume though.

Re: Emulating vacuum tube buffer in DSP?

Well, the effect is subtle! It behaves as a soft-clipper!

Here is a short article, if you want to know a bit more:

https://soundista.com/soft-clipping-explained/

Re: Emulating vacuum tube buffer in DSP?

MetalHeadKeys wrote:

Well, the effect is subtle! It behaves as a soft-clipper!

Here is a short article, if you want to know a bit more:

https://soundista.com/soft-clipping-explained/

That's interesting, but wouldn't it by definition of "clipping" mean that characteristic of sound will change with volume?

18 (edited by KaiS 2020-03-30 22:46:11)

Re: Emulating vacuum tube buffer in DSP?

Sebastian.Athea wrote:
MetalHeadKeys wrote:

Well, the effect is subtle! It behaves as a soft-clipper!

Here is a short article, if you want to know a bit more:

https://soundista.com/soft-clipping-explained/

That's interesting, but wouldn't it by definition of "clipping" mean that characteristic of sound will change with volume?

Yes, in principle.

With tubes the transition from "perfectly clean" through "soft saturated" to "obviously distorted" can take a range of 15-20dB for a single sided triode amp without NFB (negative feedback).

It can even be almost similar to a typical solid state amp with strong NFB, having no transition, directly going from clean into hard clipping.
This is not very typical for tube designs, but does exist.

The clipping behavior depends on the way the tube is used in a circuit, read my 1st posting in this thread.

If heavily overdriven there is no difference in either version.
Even a tube circuit has an absolute maximum output level, once this full saturation is reached waveforms are cut, the sound gets heavily distorted.


If I could see it's schematic I could tell which type your buffer amp belongs to.
Knowing the tube type and amp's gain (difference of loudness with or without the amp) could already give a coarse qualification.