Microprocessor Bias Control For Tube Amps

[Enzo]

That is the main point in fact. "Bias" has taken on religious qualities over time, and people specify it as if it were much more critical than it is. Imagine adjusting the idle speed of your car engine. If the factory spec is 700 rpm, would we really obsess over whether it was 694 rpm instead? Or 702 rpm? We start looking at bias all by itself without considering the overall context of just a guitar amp, and we get caught up in minutia.

This is a fine exercise in engineering, but ultimately bias simply means adjusting the control grid voltage to set the operating point of the tube. There is no consensus on what that should be, only rules of thumb and traditions. So it is up to the designer to decide if a certain current level is average enough for most tubes or it it needs more complexity such as a dissipation calculation or tube selection input.

[Leo_Gnardo]

I think that's a good point, why would one waste an uP on a rather simple function as biasing when its processing power could be better utilized for something far more difficult and interesting (at least to me) - amp modeling? Just saying...

Couple years ago there was a Traynor bass amp with 4 6550's that didnt have a uP but did employ a sliding bias, transistorized.

Sounded OK 'til you started to hit clip. Then it sounded like you were making a hard landing on broken rocks. Ow...

I'll stick with the good ol' fashioned methods.

[flyingdutchman]

Couple years ago there was a Traynor bass amp with 4 6550's that didnt have a uP but did employ a sliding bias, transistorized.

Traynor, eh? Apparently they like to experiment with bias circuitry.
I've had one of their guitar combos in, it had fixed bias but with an auto-balancing circuit. A simple PNP differential pair with the low pass filtered voltage drop across the cathode current sense resistors connected to the bases, and the bias voltage for the output valves taken directly from the collectors. It worked, sort of. But the offset caused by the base currents and the low pass filter series resistance introduced considerable error.

[R.G.]

I was not aware of that from Ampeg. But is is the set of colors and ranges I came up with, without knowing the Ampeg. I'd be interested in seeing when Ampeg brought it out.

Likewise the use of cathode resistor and comparators for sensing a threshold on per-tube current. I've posted about my variant of that going back about 12 years. It's somewhere on the forum. Once again, when I came up with that I mentally filed it under "one skilled in the art". It is possible for a lawyer to use that phrase to chisel on patents, all right, but I use it to keep myself from filing. Kind of the reverse.

Every time I mention the current monitor and its (to me, obvious) extensions to latching/non-latching/resettable/indicators per tube/etc. the next post is generally from someone poo-poo-ing it. As you say, it does work and if properly set up can save your tubes rectifiers, and/or power transformer.

[bob p]

I was not aware of that from Ampeg. But is is the set of colors and ranges I came up with, without knowing the Ampeg.

I guess great minds think alike.

I don't know when Ampeg came out the the idiot light system, but I know that my SVT-2 Pro amps have them. One of them is an older USA model (from the St. Louis Music era) and one of them is a more recent Korean model (from the Loud Technologies Inc. era).

The SVT-2 Pro design is old enough that Loud discontinued it last year. The fact that my Saint Louis Music Ampeg has the bias/comparator circuitry means that the design is pretty old ... coming up on 16 years at least.

Here's a copy of the SVT-2 Pro owners manual, which explains the bias system. It has a 1998 copyright date on it, so the circuitry in the SVT-2 has to be at least that old. I'm thinking that if there was a patent, that it's got to be toast by now.

http://www.ampeg.com/pdf/svt-2pro.pdf

I have the schematics if you want to see them.

[bob p]

Every time I mention the current monitor and its (to me, obvious) extensions to latching/non-latching/resettable/indicators per tube/etc. the next post is generally from someone poo-poo-ing it. As you say, it does work and if properly set up can save your tubes rectifiers, and/or power transformer.

I like the system. A lot. It rubs me the wrong way when we're talking about the comparator system and someone who doesn't appreciate the circuit will butt-in and start yapping about how "Crate-Pegs" are pieces of garbage. The system is very well designed, and it works. It's definitely saved a lot of very expensive SVT iron from worst-case scenario failures over the years. When people start to discuss the concept of "immortalizing" an amp, it's one of the first things that I think of. To me it makes sense to use solid state circuitry in applications like this one, to spare our amps from the most disastrous of failure modes. To me that's a very forward looking design that should be included in every modern well-built tube amp.

I think we discussed this before -- if somebody marketed such a board to amp builders I would buy one for every build. Now that I think about this, we were talking about a 6-way independent bias setup for each tube. I'd like to have that for my Super Twin Reverb; independent bias of each tube takes the concept of granular control to the next level. The SVT only has adjustment for each triplet, with current sensing for each tube. Although the SVT setup doesn't have per-tube granularity for bias, it does have per-tube granularity for protection, which is really nice.

I'd be happy just to see someone marketing an inexpensive protection circuit board that monitors cathode currents and forces the amp to shut down if things get ugly.

I don't know why SLM ended up designing the comparator circuit into their amps back in the 1990s. At times I wonder if they were just a bunch of forward looking guys, who realized that this was the right way to design an amp, and built a great design that was so expensive that the market just didn't realize it's value. They built amps that we so good that they went out of business. I guess it's possible that Loud Tech continues to put these designs into the current line of SVT because they prevent expensive warranty claims. Given how cheap a stuffed board of IC turns out to be, it would make sense for the manufacturer to protect the iron, and their parts department from expensive warranty claims.

I think it's a good idea either way.

[Chuck H]

I would think that monitoring amp current and auto adjusting would be more critical for failure related incidents. An appropriate bias isn't hard to achieve and operating current should follow suit. So monitor and auto adjust for failure situations, right? I can see where it MAY be possible to get more power from an amp if you could idealize the center operating point for a given signal level and circumstance. But how much more, really!?! Do we need this? The variables compared to the operational difficulties of such a circuit seem counter productive/intuitive at this time. Someone may eventually come up with a way to idealize it, but to what great advantage??? A few watts? In the interim there are all manor of unpredictable distortions that can occur rather than just plain clipping. Maybe I'm not trying to get ahead of the curve, but it doesn't seem worth the effort. Much less considering the limitations for this application as it applies to current technologies. Being so stringently analog, tubes experience huge anomalies in their operating parameters. I'm sorry, but asking logic circuits to keep up is kind of crazy IMHO.

[R.G.]

@bob p: Thanks. That's very much what I came up with too. I did find a hifi amp that used LEDs in a similar but not identical way when I did the initial patent search, and it was - 80s? maybe? - so I wasn't worried about patents on it as they'd have lapsed. But I didn't know the Ampeg version.

As for marketing a thing like that, I won't, but for other reasons. I'm fully employed on making effects in my "day job", so I consider it a conflict of interest with my employer to market stuff on the side. But I might update the old circuits and post them on geofex.

Actually, today, I'd use a single uC chip with some Rs, Cs, and diodes. The uCs can do many channels of A-D conversion, and the R/C/Ds would make the signal safe for the uC to read all the time. The chip would sit there always reading the filtered DC levels and when one exceeded X current for more than Y seconds, shut things down.

The only finesse involved would be exactly what "shut things down" means. To the uC, it's all just a logic signal output for on/off. It could shut down the signal to the output tubes, or the screen voltages, or the current in the tubes through a MOSFET in the cathodes, or the B+ after the filter caps, or the B+ before or after the rectifiers, or the AC power coming in to the amp. The shutdown could be until a "reset" button is pushed or require a power off/on cycle on the power switch. I used the MOSFET in the cathodes on the original version I did with the comparators and logic. But there are reasons for and against each version of shutdown that need some thinking about what's best in a musician's working setup. uCs are so fast today that all of this is a big UNDER use of the uC; it would be loafing, and spend most of its time in a loop checking to see if it was time yet to go make the rounds of checking currents.

But I've always been taken with the idea of using a lot of cheap silicon to make the world safe for fragile, expensive tubes. Silicon is cheap, and gets better and cheaper all the time. We should save the tubes for the delicate sounds they create.

@Chuck h: I don't think anyone has in mind having the bias controller tinker with the bias as you play. In fact, my early work taught me quickly to NOT mess with that bias while the speakers were not muted. It can be loud unless you can match the steps on each output tube well or do it very slowly. No, bias is set at intervals when not playing. I don't think anyone wants to use it on the fly.

Really, the red/green approach was so very fast and easy for a human to do that there was not any use for the expense of a fancy controller.

[Gregg]

I think we discussed this before -- if somebody marketed such a board to amp builders I would buy one for every build. Now that I think about this, we were talking about a 6-way independent bias setup for each tube.

In SVT2 the bias is obtained by direct coupling a CF to the grids. If you need individual bias for each tube I guess you'll need more CFs. If it's a "regular" 4 or 6 way bias it's much easier to do.

Concerning the SVT2 protection circuit I've built it in a clone of SVT2 power section and it works fine. You can find the full SVT2 schematic including the technical bulletins below|:

SVT2PRO.pdf

[Leo_Gnardo]

@bob p: Thanks. That's very much what I came up with too. I did find a hifi amp that used LEDs in a similar but not identical way when I did the initial patent search, and it was - 80s? maybe? - so I wasn't worried about patents on it as they'd have lapsed. But I didn't know the Ampeg version.

Ran across this in a Conrad Johnson power amp. Red LED lights to indicate sufficient bias. Just turn trimpot and stop turning when you see red.

For @ last 30 years, the US Patent Office seems to be out to lunch. Seems they will issue a patent on anything. Not concerned with prior art or "obvious to one experienced in the art." Here's your patent, thanks for the check, and y'all can go fight it out in court. They even issued a patent for a dog toy - a stick - since rescinded - just to show you how little attention is paid. One of my college apartment mates has a career as a US patent examiner. Perfect job for that goofball.

The only finesse involved would be exactly what "shut things down" means. To the uC, it's all just a logic signal output for on/off. It could shut down the signal to the output tubes, or the screen voltages, or the current in the tubes through a MOSFET in the cathodes, or the B+ after the filter caps, or the B+ before or after the rectifiers, or the AC power coming in to the amp. The shutdown could be until a "reset" button is pushed or require a power off/on cycle on the power switch.

Had a situation onstage 5 Sept 2005 with an SVT-CL. Stage voltage was too high and "couldn't be fixed." (The real problem: a squabble with Montreal union electricians who objected to working Labor Day without a bonus.) Band thought up a workaround, and elected to play "acoustic." Part way thru the show, the bass player tried to switch on his SVT-CL, and the pilot LED flashed red-green-red-green indicating a fault, and wouldn't power up. So it was smart enough to sense dangerous operating conditions & refuse to apply hi voltage to the output circuit. File under "The Amp That Saved Itself." Not such a bad idea to have some solid state control. When this idea first came out I bristled, but now accept it.

I don't think there was a uC in the SVT-CL, but enough logic and comparator circuitry to save the day.

[bob p]

But I might update the old circuits and post them on geofex.

I appreciate the work that you've done at GEO. Just thought I'd say that, since I use the site more often than I say thanks.

As a DIY guy, I'd like to *BUY* a turnkey uC based solution, and I have *ZERO* interest in dealing with a uC otherwise. In specific, I'm not willing to do the programming. Someone would have to sell loaded chips. My aversion to coding a uC isn't really an aversion to coding -- it's more of an averstion to the hurdles that have to be jumped through to get a uC development setup up and running. For a one-off build these hurdles are so high that I don't find uC to be a desirable solution.

The last time that I looked at a uC solution was for a timer circuit. I needed to design a flash trigger & quench controller to synchronize the firing of an array of camera strobe lights, and allow the user to manually control the duration of flash to control light intensity. Somebody had already developed a system to do something similar using a MicroChip branded uC solution. The parts count was minimal and the uC code was available, but getting a programmed uC was impossible. To load a single uC with code, I was going to have to buy an integrated development environment and a uC programmer that were both specific to that one uC. It was not only cost prohibitive, but excessively labor intensive for a one-off application. And to top it off the uC platform was obsolete by today's standards.

It took me a few days of study to realize that the uC solution was obsolete and buying specific hardware /software for that particular uC just wasn't worth the effort. I chucked the design into the garbage and designed a brain-dead solution using 55x timer IC. That really simplified things: Instead of programming buttons and an LCD display, I just used rotary knobs and RC circuits. The result was that the device had a nice analog control feel where you could just turn a knob change the light output in 1/3 f-stop equivalents. It ended up being easier to use than the uC-based interface would have been.

As far as DIY projects go, I think that people who like to design and use uC will choose a particular platform to work with, and they leverage their sunk costs by using that platform for all of their projects. That's great for the guy who designs uC based projects and can deal with a single platform, but not so good for people who might want to follow him. What inevitably happens to the "followers" is that they encounter multiple different uC based solutions on the web, and they always use a different platform. The result is that users have to deal with multiple uC platforms. The sunk costs associated with having multiple platforms tends to eliminate the appeal of uC based solutions.

I think uC controlled solutions are great when you can amortize the development cost across a large number of units, but for simple one-off projects their costs seem prohibitive.

So if you plan to update circuits at Geo, please think about using solutions that don't require investment in a uC development system. Although they make sense at the development end, at the production end they're just not an efficient implementation for low production numbers. That tends to limit their appeal to the DIY crowd who build in small numbers.

[bob p]

I don't think there was a uC in the SVT-CL, but enough logic and comparator circuitry to save the day.

Sometimes simple is better.

My understanding is that the Ampeg amps produced by LOUD continue to use the old comparator circuitry developed at SLM. (Like my SVT-2 Pro -- both use the same circuit.)

The bias circuit uses TL072 and TL074 as comparators, nothing fancy like uC. Why fix the circuit if it's not broken? Opamps are cheap, and shifting to a uC would require R&D expense without resulting in any real payoff.

An added argument for opamps is that with a uC based solution servicability becomes an issue. If an opamp dies, they're easy to service in the field. Most amp repair guys couldn't fix a failed uC, so having a uC only makes repair more complicated.

Biasing and protecting an amp are not rocket science. Sometimes simple solutions are better. I really like Ampeg's idiot light system.

[R.G.]

I understand and sympathize. Yes, uCs were once difficult to get filled up with bits. However that is no longer necessarily true. My favorite flavor of uC, the Microchips Technology PIC line can mostly be programmed in-circuit by glomming onto power, ground, and three pins of the uC. The "programmer" for this is an ordinary PC with a modified connector stuck into the printer or serial port. See:
Very Simple ICSP Programmer
If hacking a port is not your style, there a low cost programmer:
Low-Cost PIC ICSP Programmer, serial port ; that one is under $10, which is costly for one, but no one ever programs just one.

For just programming the chip if the code is available, you don't need a development system. If you do, I use the Microchips MPLAB development environment which is *free* although it does require you learning how to drive. But if you have the binary/HEX code for the chip, all you need is a programmer and something in your PC to run the programming. And those tend to be free, even if you have to go looking for them.

It sounds like a big mountain to get over, but honestly, it's a bump in the road. And - there is likely to be someone you already know or can reach on a PIC forum who would program one for you for free. I would - mail me a chip and email the program and I'll fill your chip. Takes about five minutes. The biggest hassle there is the post office issues. But that's another whole topic.

Probably what I ought to do is to post the HEX code at GEO with links to cheap/easy programmers.

Seriously - uCs used to be rocket surgery, but they have become much more like "Do What I Want" chips.

[Gregg]

Some time ago I was also reluctant to handle uCUs but learning how to load a HEX file into a PIC is not a big deal.
For example Pickit2 programmers are available for cheap from Ebay and it takes minutes to learn how to do it.

[Ted]

I didn't know what I was starting here.
Lots of good sense.
To give a little bit of context to my original request, I'm designing a Bass Amp (for my own use) with 6 GU50s driven by MOSFET source followers with toroidal transformers.

So what i'm looking for is:

i) Guaranteed DC Balance
ii) Smart handling of fault conditions
iii) The ability not to worry about drift and manufacturing tolerances in 40 year old Soviet tubes.

One of the things I was fishing for was whether any commercial system attempted to rebias "on-the-fly" at quiet points or whether there was some smart algorithm for doing this. I'm not terribly surprised to find that no-one thought that a good idea. The dumb as a bag of hammers, power up, let currents stabilise then chop in to the bias point (and probably iterate until thermal stability is achieved) approach seems like the lowest risk.

RGs point about using cheap silicon to protect expensive iron and glass is exactly where I was heading. (Not that GU50s are even that expensive). But the possibility of shutting down on failure without burning anything is always there. 8-bit micros are cheap. For a couple of quid and a few hours programming effort, a cheapish micro with 6 PWMS will give control of bias with a bandwidth of several hundred Hz - which is ridiculous overkill. A few spare A-D I/Ps (or some multiplexing) and you can monitor Screen Grid currents.

I'm not for one moment suggesting that all of this is necessary, but it has been my experience in the past that the more you instrument a system, the more you learn about its behaviour. And learning is a good thing right?

One thing that I tend to do when designing microcontroller-based systems is to overdesign the hardware. For instance, suppose that I design my hardware with individual control of biasing for 6 O/P valves. If setting all 6 to the same bias gives an OK result then there's no need to go any further. If its not good enough then there's the possibility to drive them as two triplets. If thats not good enough then individual control.

So far as "why do it" - it's simple and it gives a more flexible solution than analog control. Almost by definition, an amp that you build for yourself is a test bed for ideas, to be played with and experimented on. I like the idea of being able to rebias a sextet of big tubes to whatever current I choose, just to see what they sound like. (I don't have any illusions about a search for ultimate tone, or indeed that anything but gross changes in bias will cause perceptible changes). As for using a micro for "more interesting" functions like amp modelling - good luck doing that with an 8-bit micro.

And I would question the sanity of anyone who thought that building a Champ with uC bias control was a sensible idea, although I guess that my previous argument about instrumentation and learning would still apply.

The point about patent law is taken. I don't think I'm imagining this, I seem to remember a few years back Hewlett Packard attempted to patent the concept of interrupts...