• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

The Midlife Crisis - My 833C Amp Build

The machining of the chassis is done.

All chassis parts save the faceplates have been sent off to Landfall Systems to be anodized smooth black (non-brushed finish). I had originally intended to have the top plate chromed to match my preamp but in thinking about it there is going to be so much light given off by the 883s that the last thing I need is a big reflector under them, so black they will be. That ought to help with heat dissipation from the top plate, as well.

Once the anodized parts come back, final assembly can begin!
 
Once the amp goes into A2, however, the distortion spectrum changes, with much less 2H and more 3H. The second picture is the spectrum at 84W output into 4 ohms. Distortion is now predominantly 3H, with the 3H at .76%. This is what was seen in all analyses I ran with the amp in A2.

A2 distortion and it's spectra are extremely dependant on the driver. The lower it's output impedance the better. Most A2 amps are transformer coupled, but I've had better results using a very high gm cathode follower directly driving the power tube.

Anecdotally, I've read reports that A2 sounds a little edgier and a little brighter than A1, and those FFT results seem to reinforce that. In any case, most listening will be in A1, except for crescendos and perhaps drunken volume cranking...
I've found that my A2 amps almost never actually transition to A2 unless I am playing it ridiculously loud, or into an inefficient speaker.


Note: in order to get consistent, believable results from this LTSpice analysis, I had to disable all compression, force double precision math, and set max timestep to 1uS. Otherwise, there's a lot of hash and false peaks along the baseline.
Where is the force double precision option? I looked and was unable to find it...

Beware that the simulation is only as good as the models. If your tube model accurately models the tube characteristics and grid current in A2 mode, then you probably get good results. Personally, I've never been able to get accurate prediction of the THD of a tube circuit in LTspice. Especially when the circuit behavior near clipping (or A2) is simulated.

For DC and small-signal AC sims, I generally get a very good match between sims and reality.

~Tom

What we need is to get some talented folks together and hack some native, full fledged modelling code into an open source simulator like ngspice. Something like a vacuum tube version of the BSIM or Ebers-Moll models for semiconductors. It would be a daunting undertaking, but the results would be worth it.
 
Last edited:
I've found that my A2 amps almost never actually transition to A2 unless

I built an 845SE that makes 40 WPC in A2. It stays in A1 until 25 watts. A2 is only seen on peaks.....the same peaks where the speaker cone is nearing Xmax. You won't hear the amp's distortion, which on my design is still mostly 2H.

LT spice is good, but vacuum tube models where grid current is involved are usually not so accurate.
 
Where is the force double precision option? I looked and was unable to find it...

.OPTIONS numdgt=7. Any number more than 6 will force LTSPICE to switch to double precision math on dependent variables.



In fact, here's a little webpage I ran across that talks about doing distortion predictions with LTSpice. Some nice little tidbits in there - notice how when he uses the default settings even his pure voltage source looks like crap!

Distortion measurements with LTSPICE | Audio PerfectionAudio Perfection

.
 
I built an 845SE that makes 40 WPC in A2. It stays in A1 until 25 watts. A2 is only seen on peaks.....the same peaks where the speaker cone is nearing Xmax. You won't hear the amp's distortion, which on my design is still mostly 2H.

LT spice is good, but vacuum tube models where grid current is involved are usually not so accurate.

The reason I suggested ngspice is that it is open source. Vacuum tube modelling could be coded directly into the simulation engine, which would allow control of parameters that are very difficult to do (accurately) with the current method of modelling.

.OPTIONS numdgt=7. Any number more than 6 will force LTSPICE to switch to double precision math on dependent variables.



In fact, here's a little webpage I ran across that talks about doing distortion predictions with LTSpice. Some nice little tidbits in there - notice how when he uses the default settings even his pure voltage source looks like crap!

Distortion measurements with LTSPICE | Audio PerfectionAudio Perfection

.

Cool, thanks for the info!
 
Hi James,

Using the clips for 6.3mm (diameter), they fit for 8mm also...
and selfmade holder.

some Pictures

An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.



An externally hosted image should be here but it was not working when we last tested it.


Best regards, Matthias
 
While waiting for the chassis to be anodized I made up a pair of mounts/heatsinks for the Rod Coleman filament regulators. They are built from a 10mm thick 3.5"x 4" aluminum block, with a salvaged heat sink from an old subwoofer amp screwed on. The filament regulators screw to the side of the block opposite the heat sinks. The blocks are drilled and tapped to mount snugly to the underside of the chassis top plate, so between the block, the heat sink and the top plate I ought to have enough thermal dissipation (hopefully, since I don't know the degC/W of the salvaged heat sinks).
 

Attachments

  • 833 Amp Filament Heat Sinks.jpg
    833 Amp Filament Heat Sinks.jpg
    21.6 KB · Views: 957
The heat sinks look to be about 10x15x5 cm or so. They'll probably have a thermal resistance around 1.5 º/W if mounted with the fins vertically. That's my guess based on similar heat sinks.

If you really care, you can always measure the thermal resistance.

Don't forget to use thermal goop between the top plate and the heat sink and aluminum block. Thermal sandwich... :)

~Tom
 
The heat sinks look to be about 10x15x5 cm or so. They'll probably have a thermal resistance around 1.5 º/W if mounted with the fins vertically. That's my guess based on similar heat sinks.

If you really care, you can always measure the thermal resistance.

Don't forget to use thermal goop between the top plate and the heat sink and aluminum block. Thermal sandwich... :)

~Tom

The sinks are 11x6.5x3cm, so a little smaller. They are also located near the fan intake so hopefully the 33cfm of the fan will boost their heat dissipation somewhat. I used boron nitride heat sink grease between the sinks and the block, and will repeat that at all locations with thermal interfaces. I also plan to use an aluminum clamp to hold down the transistors so that ought to add some dissipation from the other side of the chip as well.
 
The vast majority of the heat will flow through the metal tab of the package as the thermal resistance through that path is much lower than through the plastic package. Unless the clamps make assembly easier, I would probably stick with the normal screw + shoulder mounting method. If you want to get really fancy, you get a torque screwdriver and tighten the screws to the torque recommended by the manufacturer of the thermal washers.

~Tom
 
There is an additional wrinkle here, in that the Coleman regulators will be floated at approximately 250VDC in order to provide the appropriate bias. Rod Coleman himself advised me not to use a metal mounting screw with shoulder washer in order to avoid a potential failure. The way I look at it, I have a couple choices:

1) Ignore Rod's advice. I'm inclined not to.
2) Use a nylon screw. Since nylon doesn't melt until 220C, and the chip would be toast well below that point, it should work...unless it comes loose over repeated heating/cooling cycles. Loctite might help there.
3) Use a clamping bar across the chip package, and maybe a nylon screw through the TO-220 tab as well.

I'm leaning toward #3.
 
There is an additional wrinkle here, in that the Coleman regulators will be floated at approximately 250VDC in order to provide the appropriate bias. Rod Coleman himself advised me not to use a metal mounting screw with shoulder washer in order to avoid a potential failure. The way I look at it, I have a couple choices:

1) Ignore Rod's advice. I'm inclined not to.
2) Use a nylon screw. Since nylon doesn't melt until 220C, and the chip would be toast well below that point, it should work...unless it comes loose over repeated heating/cooling cycles. Loctite might help there.
3) Use a clamping bar across the chip package, and maybe a nylon screw through the TO-220 tab as well.

I'm leaning toward #3.



im no expert at all but, 3 gets my vote.

cant go wrong with overkill.
 
Hi Magz, you could also use ceramic insulators under the TO220 packages:

Insulators - heat sink interface material

You can also get longer and slightly thicker insulating bushes for the screws. These need the top of the threaded hole in the heatsink counter boring slightly to allow the bush to sit down on the tab.

I have used these before on floating HT regulators and they work fine, certainly better than mica or silicone insulators. i am not at home at the moment but I can post a photo later if you like.

Cheers
Matt.
 
There is an additional wrinkle here, in that the Coleman regulators will be floated at approximately 250VDC in order to provide the appropriate bias. Rod Coleman himself advised me not to use a metal mounting screw with shoulder washer in order to avoid a potential failure. The way I look at it, I have a couple choices:

1) Ignore Rod's advice. I'm inclined not to.
2) Use a nylon screw. Since nylon doesn't melt until 220C, and the chip would be toast well below that point, it should work...unless it comes loose over repeated heating/cooling cycles. Loctite might help there.
3) Use a clamping bar across the chip package, and maybe a nylon screw through the TO-220 tab as well.

I'm leaning toward #3.

It is possible to use the screw-fixing method across 250V, but a high degree of care is needed in cleaning up the hole (dia 3.5mm please, be sure to allow the shoulder-washer to enter the hole). The finish and flatness of the surface must be carefully inspected.

other considerations:

- the 833 cold filament is probably near zero ohms at turn-ON. The Coleman Regulator soft-starts the filament with a gentle ramp-up of voltage, but the startup interval still applies heightened thermal stress on Q5. The thermal interface must be in good shape!

- The materials of the heatsink, transistor, screw and shoulder washer are all different, so differential expansion may be at work, and a flash-test on a cold assembly does not guarantee that warmed-up, the insulation integrity will be maintained.

- screw method is not preferred in any event. It applies the clamping force to one end of the TO-220, unlike clips and clamps acting on the insulated portion of the transistor body.

- spring clips are purpose-designed to apply the correct mounting force to the body of the TO-220, and are superior to screws in this respect. In addition, there is no insulation weakness at the screw hole.

- for this purpose, I like the Aavid heatsink clips, or the Laird Technologies version that can be screw-mounted next to the TO-220:

TSC607-ZP - LAIRD TECHNOLOGIES - CLIP, 2 X, TO-220 | Farnell United Kingdom

there are many other similar types.

If these are used, there is no need to use any TO-220 screws, and a plain Silicone pad should be used in place of the Mica insulators I supplied. Thermal grease is not usually necessary with sil-pads, provided the interface surface is smooth and carefully cleaned.

3223-07FR-43 - BERGQUIST - THERMAL PAD, TO-220 | Newark

- if the heatsink spreader-bar is hard-anodised, there will be extra insulation safety (anodised heatsinks often present an insulating surface - it's the oxide!).


The only other precaution is to monitor the transistor body temperature during first-time use, using an IR gun-type instrument.