VVR implementation on AB763 Deluxe reverb

[Skydragon]

Starting this thread, to discuss this topic and also to help others looking at adding VVR voltage scaling to a AB763 type amp.

As background - I love the sound of an overdriven and cranked-up Deluxe Reverb, but I can't always make that much noise! so, wanted to build an amp that could also be used at low volumes. Accepting that part of that 'cranked sound' comes from the speaker, the OT and sheer movement of air, I knew I wasn't going to get the same tone at low volumes, as I would at high volumes, but I wanted to try and get as close as I could to that chimey, bell-like- tone with the gritty overdriven distortion on top. Think SRV.

I'd already added VVR to a 5e3 Tweed Deluxe clone I'd built, so using that knowledge and a bit of research I built a single-channel non-vibrato AV763 amp, as an amp head and using a 12" jensen P12Q speaker for testing.

I'll post the schematics and details shortly (I'm currently travelling) but the amp is an exact copy of the original AB763 circuit, vibrato channel only, but with the vibrato circuit removed and with a master volume added just before the phase invertor tube. I've used a new Princeton chassis as the basis of the build.

Since the majority of distortion in a DR seems to come from the power output stage and in the tests I did, I found clipping the phase invertor sounded really horrible, I decided to apply VVR voltage scaling to the 6V6 output stage only (not to the preamp or phase invertor stages). The idea being to drive the output stage with a fairly clean signal controlled by a master volume and reduce the plate voltages down on the 6V6 to the point they clipped.

I then added a two-part VVR circuit (schematic to follow) to control the B+ to the 6V6 tubes only and in parallel to track the fixed bias as the B+ reduces.

My first tests sounded Ok, but nothing like the tone I was looking for. I could wind down the voltage on the 6V6 tubes to around 110v to get a low volume and then increase the master volume until the output stage was just being pushed to clipping. However the resulting tone was poor. The distortion sounded really harsh and nothing like the warm tube tone I hoped for.

To cut a long story short, I tried to find some info about the correct bias point when voltage scaling a 6V6 output stage, but i couldn't find any info. So, I then messed around with the bias point on the 6V6. Most VVR circuits I'd seen online, reduced the 6V6 bias to a point that when scaled to 100v or so, the 6v6 tubes would only be drawing a few mA of current. I experimented with this and found that just like the bias needs to set at around 20-25ma at normal voltage levels to get a good warm tone, then the exact same applied to VVR usage. I ended up trying a bias current of 20ma per 6V6 tube and suddenly  the tone was amazing and exactly what I was looking for.

The drawback with this is the mosfet in the VVR circuit is now having to drop around 340v across it at 20ma per tube, which equates to approx 14 watts of power at idle...that is dumping a whole load of heat! which means a heatsink is needed.

I'm wondering if some VVR circuits intentionally bias the 6V6 cold so as to avoid this power/heat issue?

Edited to be more accurate >> The main reason I'm posting this thread, is that I'm guessing there may have been people who've built fixed bias amps with VVR, not realising that the bias current needed to be set as 'hot' as normal (non VVR) operation? If I hadn't messed around with the 6v6 bias point at low plate voltages, i would have still used the amp, but just accepted it didn't sound great at low volumes.

Anyway, the end result was I added a copper heatsink plate to the chassis (one inside and one outside, connected via some copper rivets to dissipate the heat). This approach is good for a self build amp, but might be difficult to retrofit in a shop bought amp (?)

Anyone any similar experiences or info to add?

More info/photos/schematics of amp and VVR to follow

[PRR]

> just like the bias needs to set at around 20-25ma at normal voltage levels to get a good warm tone, then the exact same applied to VVR usage.

Good point.

> a whole load of heat! which means a heatsink is needed.

So put on a heatsink.

> might be difficult to retrofit in a shop bought amp

Retrofits are often tough. It would be tough to put a Chevy 502 V-8 in my Honda sedan (134cid).

Occasionally new technology helps. Nitrous, dual-blowers and ice-chest "can" get as much power in the Honda engine as a sane 502 V-8. (For 7 seconds, no passenger seat.) Switch-mode power supplies can drop 400V to 100V at 90% efficiency, couple-Watt instead of 14W. But design is non-trivial, demand for this size is tiny, so few good choices.

[sluckey]

Maybe upgrade that copper heatsink with an aluminum heatsink.

[2deaf]

You're not really using the electronic terms correctly, but I think I know what you mean.  You are actually biasing the the tubes colder and colder dissipation wise when you decrease the voltage but hold the current constant.  If you tried to bias the tubes at the same dissipation (say 70% of max., considered "hot"), you would have to increase the current as you decrease the voltage.  At some point around 150V to 175V, the bias voltage reaches zero in order to get 70% dissipation out of a typical 6V6.

In addition to subjective sonic tests, I ran some objective tests with decreasing plate voltages.  One of them was to set the idle bias so that crossover distortion first starts to appear when the signal is applied to the point of output clipping.  From 375V to 250V, a steady change toward zero in the bias voltage consistently put the cathode current at close to 22mA for the same crossover distortion condition.  Below 250V, it took more idle cathode current to get the same crossover condition.  At 100V, it took an idle cathode current of 32mA to get the same crossover condition.  So a linear decrease in plate voltage required an exponential increase in idle current for the same results.  The bias voltage, however, became more positive in a reasonable linear fashion.  There is also a lot of variation between 6V6's, but the general pattern is the same.

 

 

[Skydragon]

Thanks for all the feedback and info, much appreciated.

You are actually biasing the the tubes colder and colder dissipation wise when you decrease the voltage but hold the current constant

yes, agreed, in terms of less power dissipation. As you show in your next comment about biasing to get similar crossover distortion levels, the bias current can remain fairly static as plate voltage drops (to a point anyway), with the advantage (hopefully) of less stress on tubes and a longer life when being run at low VVR power settings.

Below 250V, it took more idle cathode current to get the same crossover condition.  At 100V, it took an idle cathode current of 32mA to get the same crossover condition.  So a linear decrease in plate voltage required an exponential increase in idle current for the same results.  The bias voltage, however, became more positive in a reasonable linear fashion

That is really useful and interesting info, thank you. I've not yet tried setting the idle bias current above 25ma at <150v plate voltages, purely because I was already pushing the thermal limits of the mosfet in the VVR. Now I've better heatsinking fitted, I think I'll try increasing the bias current it to the levels you mention and see if I can hear a difference.

Maybe upgrade that copper heatsink with an aluminum heatsink.

I used copper to make a series of heatsink plates, to transfer the heat away from the mosfets, to the outside of the steel chassis (as copper is a great conductor of heat). If I start running into thermal problems, I'll attach aluminium heatsink to the outside copper plate to give better thermal radiation/

[PRR]

> upgrade that copper heatsink with an aluminum heatsink.

For size, Copper is a bit better than Aluminum.

For COST, Aluminum is much cheaper. And die-to-air transfer is mostly about how BIG the outside is. Aluminum is much better for area when cost or weight is limited.

This is of course known to the transistor boys. The power devices are historically on a Copper thermal pad, inside the package, to spread-out from the teeny die to a larger area of the package. At about that point it makes sense to switch to Aluminum.

"Scaling": two thoughts. Keep V/I ratio constant as V drops. Makes sense. Except Gm falls off with I. You can easily get into a zone of gross crossover distortion. Keeping I fairly steady gives a fairly constant ratio of load impedance to Gm, which may really be what we want. This does double the max heat in the pass device.

[Ritchie200]

Any way to get that mosfet outside the chassis on top of a finned heat sink? An enclosed chassis might get a little toasty inside. Also get more air moving across it.
Jim

[dude]

I've had heat issues with VVR's, causing premature failure.


I how use a PC's small 2" fan and heatsink, if you can find an old PC with a 2" square fan and heatsink (goes over the processor), put it on the outside of the chassis over the VVR. These heatsinks are usually 2" sq. by about 1" to 2" in height with fan on top. Tap voltage off the filament supply with a small bridge R. You'll get about 5 vdc to 6vdc, to the fan, just enough to create a small breeze but can't be heard. Works like a charm, I never had a blown VVR since I started using these old PC processor heat sinks. If interested, I can post a schematic of the voltage supply from the filament, several ways can be done.


al

[Skydragon]

Good comments and feedback. Thanks.

After seeing the mosfet regulating the B+ rail to the power tubes getting really hot when I first built the amp and realising that even if they worked, they probably wouldn't last long. I took the good old sledgehammer to crack a nut approach... 

The mosfets are bolted to a 3" long piece of 1/8" thick copper bar inside the amp, which is then connected to a longer copper bar that runs along the outside of the amp (the two pieces of copper bar are connected together through the steel chassis by some copper rivets and a few steel bolts, with all parts coated in heatsink grease). Last but not least I've connected a large aluminium heatsink to the outside copper bar. This is a 'heat-pipe' approach where you rely on the copper to transmit heat from a device to an external heatsink - fwiw a company I used to work for took this approach to passively cool high-powered processors without having to use cooling fans. it was a more expensive approach but avoided fan noise issues and had a better long-term reliability factor, when compared to a fan, which can fail.

The heat is transmitted away from the mosfet really well by the copper bar, the result is even when the main mosfet is having to dissipate >22W of heat, the mosfet casing itself doesn't get hotter than 110 - 120 degF (measured with an infrared thermometer )

A smaller heatsink and a fan would probably work just as well. But I took this approach, as I have a loathing of fans failing, which in this day and age is probably unfounded.

Taking 2Deaf's feedback from his post above, I've been messing around with the grid bias on the VVR and the resulting cathode idle current at low plate voltages. I'm now applying the 'normal' 22ma per 6V6 at 450v, but have that increasing to circa 35ma at 100v. That extra cathode current at lower voltages definitely produces a warmer tone (than letting the cathode current drop to very low values at low plate voltages).

The result sounds really good. For in-home (quiet) usage, I can set the master volume at a low level, to get the low sound level I want and then wind the B+ down to the 6V6 power tubes, down to the point I get the distortion level I'm looking for. It works really well and is easy to get a lot of really smooth, warm sounding overdriven tones from the 6v6's. For 'clean' sounds at low volume, I generally don't use the VVR and run the amp at normal voltages.

The bias doesn't track as well as I'd like on the last part of the VVR adjustment, with the cathode idle current having a peak of 40ma at around 120v and then dropping to around 20mA at 80V, which is slightly irritating, but that's due to the simple VVR circuit I'm using and I'm guessing the non-linear nature of the 6V6 at that point. it's a limitation I can accept.

I'll post some photos and schematics soon.

[Skydragon]

Here are the two schematics I promised.

The first for the amp itself

The second for the VVR circuit I'm using in this amp