Kits & Books

Search the Forum

(Advanced Search)

  Thermal grease?
Posted by: K O'Connor - 10-17-2018, 10:55 PM - Forum: Power Scaling - Replies (1)

Hi Guys

It is a common practice to use silicone thermal grease when mounting power semiconductors on a chassis or heatsink. The grease itself is fine to use except that it is universally NOT APPLIED CORRECTLY. The grease is designed to fill MICRO voids in the surfaces of the semiconductor mounting surface, the insulator, and the surface to be mounted on. A tiny amount of grease would be applied to the surface then scraped off with a straight edge. When applied correctly, you should not be able to see any grease without a magnifying glass or possibly a microscope!

Does anyone use thermal grease that way?

NO. No one does. You always see huge globs of the white stuff all around the devices and this means much more has been used than should have been. The reason this is unacceptable is not merely because of the permanent mess; rather, the problem is that the grease compresses over a short time and then the device is loose. Torquing the devices carefully makes no difference to this situation, it just means you are fooling yourself that you are being careful. Unfortunately, the care you are taking not to warp the mounting planes of the device or the heat sink is swamped by the inevitable loosening of the mounting and the loss of heat sinking that results, ultimately leading to device failures.

Because if this, we recommend NEVER USE THERMAL GREASE.

Way back when I started building solid-state amps, I bought a little tube of thermal grease. I was serious about building these things for life. Then I read an article in the Hewlett-Packard Journal (which the engineers where I worked received courtesy of the employer) and went home and threw away my tub of grease.

In our Power Scaling kits, the mosfets are bolted to the chassis for heat sinking. Thermal pads are supplied rather than mica insulators as the pads travel better. Even were we still using grease with mica we would not use it with the pads as the pads are designed to compress during the initial mounting. Ideally, if you unbolt the device for some reason or replace it, a new pad should be installed.

Print this item

  Two Chassis Amp - Wiring between them.
Posted by: NGW - 10-15-2018, 11:52 PM - Forum: Wiring - Replies (3)

Hi Kevin and fellow TUT followers,

I am building a large combo for a local guitarist; Push-Pull 2 x 6CA7, 3 x 10" speakers.
Due to the large cab and wanting to keep the pre valves out of the speaker box, I have built it with two aluminium chassis (like some old amps).
One, in a separate compartment above the speakers, housing the Controls, Preamp and the PI. 
The other, on the floor of the cab, housing the Power Supply and Power Amp (the valves are shielded from the speakers).
This two chassis build is new to me and not something that I have done before now.

This build requires approximately 700mm of wiring/cabling between the two chassis for the;
HT for the PI & Pre (PT: 290HX = 356VAC at 420mA), Heaters, Chassis Ground, Speaker Ground, Feedback wire (circuitry is with the PI) and Signal from PI to PA.
I am going to use a relay, via a control panel switch, to switch the wall supply to the PS.

I have some ideas about which way to do this, but I thought I might ask what you blokes recommend, so I can get the best result, with no oscillation or noise and to ensure it is safe.
All suggestions gratefully received.

Cheers, Noel

Print this item

  Fuse holder, Fuse Clips and Wiring
Posted by: makinrose - 10-13-2018, 04:29 PM - Forum: Wiring - Replies (2)

The TUT series recommends a lot more fusing than most builders use for MI tube amps but it make sense to me to provide the maximum protection possible to the transformers and thus also protecting  the owner from expensive replacements. Up until now I've been using two panel mounted fuse holders (one for the mains and on on the HT). 

I've been trying to work out the best method or part to use to mount fuse clips/holder to the eyelet boards so I can add more fusing.  What method do you guys use?  Any recommended fuse holder/clips/fuse blocks that are easy to mount on a eyelet board?   Is it preferable to still have some of the fuses be panel mounted or better to have them all be internal.  Thanks!

Print this item

  Frequency response using white noise and FFT
Posted by: K O'Connor - 10-09-2018, 02:44 PM - Forum: Test & Measurement - No Replies

White noise is like white light, where all the frequencies are present are once. The noise you hear between radio stations when the tuner has not found a station is white noise and the amplitude of the noise is constant with frequency. This makes it a convenient signal to use to perform a frequency response test on a circuit if IF you have a spectrum analyser.

Spectrum analysers are a specialised oscilloscope with multiple frequency filters at the input. Typically, a single frequency is fed into an amplifier, say, but the amplifier adds distortion harmonics that were not present in the original signal. The analyser shows amplitude on the vertical axis (Y) and frequency on the horizontal axis (X). t each frequency, a vertical line is displayed whose amplitude corresponds to the amplitude of that frequency. In a typical display, the original signal is the "fundamental", or "first harmonic", and is used for the reference level; ideally it has the highest amplitude. The harmonics have much lower amplitudes, so their lines are not as tall as for the fundamental. 

The vertical scale is often in deciBels, but can be in volts. The dB scale is more convenient if the amplifier is very good, exhibiting very low distortion, as the "linear appearance" of the logarithmic dB scale allows the low levels of the harmonics to be visible.Were the scale truly linear, the harmonics might be buried in the noise of the zero line across the bottom of the display. The displayed output is also called a "fast Fourier transform" (FFT), which is a mathematical derivation that would result in the same graphical output.

So, the spectrum analyser is great at showing harmonics of a single frequency input. if we feed white noise into the amplifier we see a displayed line that represents the frequency response of the amplifier. This is also called a "Bode" plot. Spectrum analysers are a bit expensive for most labs to have and unless they use it constantly it is difficult to justify the expense. Fortunately, modern DSOs (digital storage oscilloscopes) have a computer core and computers are good at doing any math the designer wishes to have it do. As a result, most DSOs will do FFTs - maybe not as well as a dedicated spectrum analyser, but well enough for most techs and hobbyists.

The only DSOs I found that specifically lists being able to display Bode plots are from Keysight (formerly Hewlett-Packard, HP - not the brown sauce from England... mmmm), EDU1002G - 50MHz, ~$900cdn and DSO1102G - 70/100MHz; ~$1200cdn. Both have abuilt-in 20MHz arbitrary wave generator (AWG) allowing the scope and generator to be perfectly synchronised to do a frequency sweep. Keysight isavailble through Mouser; HP sauce is availble at your local grocery store.

Print this item

  Frequency response using 2-ch Scope
Posted by: K O'Connor - 10-09-2018, 02:11 PM - Forum: Test & Measurement - No Replies

If you have an oscilloscope and a sine generator but the generator output amplitude is not will regulated, you can still do frequency response tests if the scope has two input channels. The key to this is simply knowing that when the 3dB-down point is reached, the phase shift of the output signal compared to the input signal will be 45-degrees.

Connect the sinewave generator to the amplifier input and connect one probe from the scope to the amplifier input, as well. Connect the bench load to the amplifier output along with the second scope probe. It does not matter what the gain of the amplifier is, just set the channel sensitivity for the scope to be appropriate for the end of the amplifier that channel is monitoring. ideally, both waves should be as close to full-scale on the display as possible.

Turn on the amp and see the waves be essentially overlapped at low-frequencies. As you sweep the frequency of the generator higher, the waves will eventually begin to spread apart.

If the scope has an X-Y function, then you can generate a Lissajous figure, commonly exploited for RF work. The problem with this technique is that the shape of the resulting ellipse depends on BOTH the amplitude of the two waves and their phase difference, so it works best if the generator output is highly stable, which we assumed was questionable at the beginning oft his discussion. if it is stable, checking output amplitude of the amplifier is much simpler, as described in another post here. The other problem with the Lissajous method is that when 45-degrees phase difference is achieved, the ellipse is a very specific ratio of diameters rather than being, say a straight line, as it is for 0-degrees or 180, or as a perfect circle at 90-degrees and 270.

Print this item

  Importance of Primary Impedance Output Transformer Matching?
Posted by: Bigjim63 - 10-08-2018, 12:29 AM - Forum: TUT Q&A - Replies (3)

I have all of the TUT books but am still confused when it comes to matching Primary Impedance on Output Transformers to suit the particular tubes being used.  Furthermore, how important is it and what difference does it make?  

In regard to the construction projects in TUT 3 & 5 the most common specified output transformer for 50 watt power amps is a Hammond 1650KA.  This is being driven from a 4x octal tube complement.  The transformer primary impedance rating is 3,400 ohms with taps for 4,8,16 ohm speaker outputs.   You can find this transformer on the Hammond website under the general heading: Push Pull "Ultra Linear" HI-FI and it specifies for use with 4 Tubes (6L6GC, 807, 5881, EL34, 6146B, 6550B).  So far all good and it seems that this transformer is perfectly matched to a 4x Tube complement.

However, if you look further on the Hammond website and look at a direct Marshall JCM800 replacement output transformer for 50 watt you get the recommended 1750N. This has a primary impedance of 3,200 ohm.  But this is driving only 2x tubes!  Why is it so low?  Shouldn't it be higher at around 5000 to 6000 ohms if driving only 2 tubes? It is interesting to note that the 100watt JCM800 replacement transformer (Model No. 1750U) driving 4 tubes has a primary impedance of 1,700 ohms.  Do Marshall deliberately mismatch their primary impedances which gives them the sound they want?  How important is it to match impedances and what is the ideal impedance for a octal tube complement?  

If anyone can answer my questions and explain this confusion I'd be very grateful.

Print this item

  Oscilloscopes for Audio
Posted by: K O'Connor - 10-04-2018, 10:11 PM - Forum: Test & Measurement - Replies (4)

Hi Guys

Being an analogue guy I've always had analogue oscilloscopes. My first one was a Philips 50MHz 2-ch that I bought new right from their warehouse for about $1500cdn. That was 1984 and it worked until about 2012 or so. I used a borrowed scope for a while, but it was old and temperamental, then bought a new GW Instek GOS-620 which is 20MHz 2-ch with a simple front panel. It worked for 18-months and then the vertical positioning went awry. A sine wave in the bottom 2cm of the display would look okay, but if you moved it upwards using the vertical position control, the top of the sine would distort, then flatten to zero a centimetre before the top of the screen. There was no way to accurately look at large waves or complex waves.

When I bought the Instek, it was about $560cdn and i looked at DSOs, as well. A DSO is a digital storage oscilloscope. They were initially designed and optimised for looking at digital systems where wave forms were mostly square waves and strings of pulses. None of that requires much vertical resolution, which was fortunate because at the time DSO were being born, analog-to-digital-converters (ADCs) were expensive to make. ADCs got better with time, faster, stronger - not bionic, though! - and are at the heart of 90% of all DSO still made. But now, you can get bandwidth out to 100s of GHz and memory depth unheard of a decade or so ago. 

Initially, DSOs had a latency issue, where it would take a short amount of time to build up the waveform before the scope would display it. Modern computer technology is blindingly fast and the latency is pretty much gone The only time you would have an issue is if you set the time base to be many 10s or 100s of seconds - something you could not do with an analogue scope - so, nothing to worry about for what we look at in guitar and hifi amps. I was concerned about that latency a year or so ago when I decided to go with the Instek CRT scope. I did not know what many of the DSO specs meant until the Instek partially died and I had to re-investigate scopes and figured I'd get a DSO this time.

Print this item

  Library parts
Posted by: K O'Connor - 09-23-2018, 03:12 PM - Forum: PCB Design - Replies (11)

Hi guys

In Eagle and other PCB layout software, the components are stored in a parts "library". Eagle comes with some libraries developed over the years by its developers, and to which you can add your own parts. It is your choice whether to use the default libraries or to make your own for things that you use all the time, as this can save a lot of scrolling through the existing libraries.

Note that the libraries are listed alphabetically in the library directory. If you create your own directory it is convenient if it shows up at the top of the list. To force this, use numbers at least as the first part of the library name, or as the entire name. I use a couple of libraries named this way: one for laying out schematics, and another for laying out mechanical drawings for chassis and face plates.

Each component is called a DEVICE. A DEVICE is comprised of a SYMBOL that goes on the schematic and a PACKAGE that goes on the board. Each device can have s many packages as you wish, and some devices might actually require more than one symbol to be complete. For example, a resistor has only one functional element and thus has only one symbol, but because the resistor can be many different physical sizes, it can have many packages. A dual-triode, like a 12AX7, has three functional elements: two triodes and a heater. If the DUAL-TRIODE device is made correctly to assure proper schematic flow, the DEVICE will have three symbols and then one or more packages as required. Ideally, the 12AX7 should have two packages: one for mounting on the top side of the board and one for mounting on the bottom side. Alternatively, you can have one package and use the MIRROR tool once the package is placed on the board.

You can go in to the library directly, but it is usual to have a schematic open before invoking the Library directory.

Print this item

  Incremental vs. Complete Schematic Capture
Posted by: K O'Connor - 09-23-2018, 02:39 PM - Forum: PCB Design - Replies (1)

When you are considering how to lay out a new project, you often visualise the finished product and work backwards. You actually envision ALL aspects of the product at once and can see the shape of the chassis, where the major components will be positioned on the chassis and on the PCB, where the controls and jacks will be, how the wiring might flow. For the purpose of this discussion, we are assuming a PCB will be the main component carrier. Using a PCB means the design work is "front-loaded" and that making a second, or third, or one-hundredth copy of the product will be very quick to do. if we were hand-wiring, every unit would require the same amount of work even if we are following a fairly specific layout. The wiring time for the PCB amp is usually much shorter than for the hand-wired amp.

Note that on the scale we are working on, ALL the work is by hand: hand stuffing of the PCB: hand soldering; hand assembly; hand wiring of whatever requires wiring; hand testing.

Suppose the project is a simple guitar amplifier.

If the amp is solid-state, then there are not too many thermal considerations other than where the output devices might be bolted to the chassis, or to a heatsink, and if the heatsinking will be adequate?

If the amp uses vacuum tubes, then we have to assure air flow around the tubes and try to keep their heat away from the PCB. Do we put the tubes on the PCB? Do we use chassis-mounted tube sockets and wire to the board from there? Let's suppose we decide to mount the tubes on the PCB. This has many advantages in that the wiring around the tube socket is "locked in" and parasitic capacitance will be controlled and will be consistent for every copy of this amplifier. Besides, card-mounted sockets eliminates nine connections per 12A_7 and up to six connections for each output tube. Suppose, then that the choice is to use PCB-mounting for the tubes.

In our vision of the amp, we know the general placement of the controls and tubes and jacks. In our modern time, the instinct would be to mount all of these items on one PCB to eliminate as much wiring as possible. Indeed, for a preamp-only or a power-amp-only this is easy to do and does not necessarily incur a weighty service penalty. For a complete amplifier it may be more problematic, so we might have to split the single large board into smaller pieces to make servicing easier.

Splitting the PCB into smaller parts also has the advantage of allowing a modular approach to the development of a product line - a "family" of amplifiers - and/or for easy construction of custom amps. It also allows one to use the less-expensive versions of Eagle or other PCB design software, where the board size is restricted.

Once we make these decisions regarding the board size and what is on it, we should first make a hand drawing with the dimensions for spacing of the externally visible and accessible components, then we can begin drawing the schematic. Because we have an idea of where some things should be placed on the PCB, we can choose to lay out the entire schematic THEN lay out the board, OR lay out a small part of the schematic and start the board, then add a few parts at a time, arranging them as we go along.

Using the entire schematic entry approach is the conventional design intent. You draw the schematic and then select the BOARD icon. Eagle asks "Create new board (named as your schematic)?". Press OK and the board editor opens with a rectangular board outline to the right and all the components arranged to the left. The parts have fly-wires between them showing what lead terminal ties to what other component terminal. It can be quite a mess and intimidating if the circuit is complex. The shapes of the components will make it obvious what each component type is, and they will be laid out in numerical order identical to the order they were dropped onto the schematic. This last detail explains why you might see R1, R2, etc at the middle or end of a schematic rather than at the beginning - the designer or draftsperson simply dropped R1 and used it in the first part of the circuit he/she was thinking about.

Note that (in older Eagle versions) the default board outline is on a 0.05" (50-mil) grid and represents the centre of the milling tool rather than the true board edge. Use the REMOVE tool to get rid of these lines, then use the DRAW tool to lay out a board outline closer to what it should really be. More on these specific actions later.

You should print out the schematic to assist with laying the board. In the board editor, you can use the MOVE command to move parts from the field onto the actual board. If you made that hand drawing of the key dimensions previously, now is when it comes in handy. You can draw reference lines on the board area to show how to precisely align and position the pots, jacks and tube sockets. Move these components first as their positions influence the positioning of all the small resistors and capacitors, and as to how to lay the traces between components.

Fortunately the MOVE feature allows you to enter the geographic designator or NAME of the component into the command line while the cursor is over the board area close to where the next part should be. Enter the name and that part is suddenly stuck to the cursor as a virtual component, ready to be dropped onto the board. left-click to drop the part, then enter the name of the next part to position. Use the ZOOM tools to better see the part of the board you are working on.

Alternatively, since you know the basic lay of the land for the board itself and likely have a schematic in mind, you might do the layout a bit at a time as you build up the schematic. Again, having the hand drawing with key dimensions for placement of the externally accessible parts is essential, although you can also make those decisions within Eagle. Suppose you open the schematic and drop the jacks and pots. Then open the board and make the proper outline for the board and add the reference lines for where the pots and jacks should go. Now move those parts from the field at left onto the board space, lining them up as they should be.

Here is where you can take advantage of the GRID command. Suppose all the pots and jacks are on a 1.5" spacing. Set the grid to 1.5" and for the grid lines to be visible. Now when you move a given component, it can only move in 1.5" increments and everything will be easily lined up to the required spacing. After these parts are in place, reset the GRID back to 0.1" or whatever small increment is appropriate. The dimensions here can just as easily be metric, with "easy" spacings based on millimetre or centimetre multiples. Using this technique might let you save the step of drawing the reference lines, but those are handy to have for other steps later, like laying out a chassis drawing.

Now you have the pots and jacks on the board where they should be and on the schematic in relative positions that may or may not promote a good schematic layout. Move them as required to make space for the next few components. Say we drop the tubes in next. 

Depending on how we have constructed our dual-triode symbol, we are either dropping individual triode sections OR dropping a dual-triode with each ADD command. Using the individually requested triode sections allows for an easier to follow schematic that does not have the signal doubling back on itself - the signal flows from left-to-right as it should. When we add the first triode section to the schematic, the entire dual-triode package is dropped into the unconnected component field in the board editor. The next triode section we drop onto the schematic does not change the board editor. The third triode section we drop adds a new duo-triode package to the board. Using the "component" dual-triode, the heater will be a third component that we drop onto the schematic later using the INVOKE tool. It takes two clicks to drop the two triodes onto the schematic, and a later step to connect the heaters, with the potential to forget about the heater wiring, but also for making a neater schematic.

If we have made the duo-triode symbol as exactly that - a dual triode with six connections - then each request for a dual-triode drops the two-section symbol on the schematic and the package onto the board. Most schematics using this symbol approach also have the heater connections within the symbol, so in fact there are nine-connection points made to the single schematic symbol. Where this is a more realistic presentation of the PHYSICAL component, it is NOT really what we want to have on the schematic. The schematic is an abstract representation of the circuit and should make following the signal and power flow EASIER than it is in the real world construction. Having this all-in-one symbol is intuitive but obfuscating at the same time. It allows each 12A_7 to be dropped onto the schematic and board in one click.

Print this item

  PCB schematic capture basics (part-1)
Posted by: K O'Connor - 09-21-2018, 01:48 AM - Forum: PCB Design - Replies (2)

Hi Guys

Using computer software to design printed circuit boards traditionally involves two steps: capturing the schematic, then laying out the board. Technically there is a third step that precedes these which is to make the library of parts, but we will assume here that a device library already exists. In which case, you could still say there is a third step afterwards, which is to generate the Gerber and Excellon files for the PCB house to use to manufacture the board.

Schematic capture is actually YOU manually drawing the schematic in the "schematic editor". 

I've only used Eagle (and Maxi-PC back in 1987) but I would think most PCB design software is more similar than different? In Eagle, you go to the File button, then select New from the drop-down menu, then select Schematic in the second drop-down menu. The schematic editor will open.

Across the top of the window are three tool bars: The top one has file management menus as File, Edit, Draw, View, Tools, Library, Options, Window, Help. Some of these functions are duplicated as icons in the second toolbar, as: folder, Save, Print, Tile, Board/schematic-jog, page sheet-number-window, Use, Script, Run, Zoom functions (full, in, out, redraw) Undo, Redo, Stop, traffic-lights. The third toolbar has just the GRID icon, which is a group of dots (?) and is very important.

For schematic capture, the GRID MUST BE SET TO 0.1" or the connections to the parts will not be made. Right click on the scale icon and the GRID window opens. You see Display on/off which provides grid-lines that are visible or not. Beside this is Style Dots/Lines, where the visible grid indication can be solid lines or lines of dots. It is normal on the schematic to turn the display OFF, in which case it does not matter which style is selected.. Below this is the Size window, where you enter 0.1 and then make sure the default "inch" is visible
Below this is a window called Multiple, which is usually set to 1 (one) and then the Alt window which might be defaulted to 0.01. These are noncritical. Press OK to set the grid.
the scale, zoom functions, 

Down the left-hand side is a tool bar with all the functions available for the schematic editor, in two columns of icons. One of them is the Add function, which looks like a logic AND gate, which is how you access the library of parts. Left-click and a window named ADD opens showing a list of libraries. Scroll through the list. If you know which library has the part you want, left-click on it and it will open and you can see all the parts in that library. Scroll through it and left-click on the part you need. When you do this, the schematic SYMBOL will appear in one of the right-hand boxes within the ADD window, and the PACKAGE will appear in the box beside it. If this is the correct part, then click on OK and the ADD window disappears and the part is now stuck to the cursor in the schematic editor.

Use the mouse to move the part anywhere on the screen you wish, then left click to drop the part. The virtual part is still attached to the cursor allowing you to drop as many of this components onto the schematic as you wish. This is handy when you need a lot of resistors, for example, or capacitors, or any other component. Each part dropped onto the schematic will be assigned a geographic NAME in the sequence of their addition, so if resistors, they will be R1, R2, R3, and so on.

You can MOVE the parts around using the MOVE icon, which is like a compass with an arrow pointing up (north). Left-click on MOVE then cursor to the part you want to move, left click on the part and it is attached to the cursor. If you right-click the part will rotate and stay on the cursor until you left-click again to drop it in place. Each rotational step is 90-degrees, which is appropriate for the schematic as all the connection points on the component, called PINS, must be on the 0.1" grid which is square. 

To connect the components to each other, we draw NETS between them. The NET icon has a hatch-mark line at top that bends and goes down to the right, with three horizontal green lines within. Left-click on this and the third tool-bar along the top will now have options for the wires that you will draw. The first are how the wires can bend, with common default being 90-degrees. The next is orthogonal (45-degrees), then random, and curving options, then the mitre-window, and infill-windows, then Style (Continuous, LongDash, ShortDash, DashDot), and the Net Class ((0-default). Generally, we use the defaults of 90-degree bends, no infill, continuous and 0-default class.

After clicking on the NET icon, we cursor to the pin of one component, then right click, then cursor up or down or sideways as appropriate. There will be a green wire attached to the cursor that will bend where we click until we reach another component pin. We can cursor to any point along the wire we have just laid and make a branching wiring to another component. A dot will appear at the junction when we make the next click on the branch, at a bend or at the next component pin.

Each group of component ends that are tied together is called a "net", short for "network". Each net is assigned a name by Eagle, as N$1, N$2, N$3, and so on. The name of the net is inconsequential to us most of the time and never see these names anyway unless we select the Information icon, or have a reason to rename the nets. We will discuss these things later.

The first time you save the schematic you will be asked where it should be saved and to name the schematic. This is mostly a computer file management issue, following the tree to a point within Eagle where the schematic should be accessible. Eagle has a Projects folder, which is where these things should go. It is a good practice to make a new project folder for your new schematic because you will eventually have a related board and the CAM files needed for manufacturing. That is twelve (12) files overall for the one board, so best to isolate it from your other projects with each in its own project folder.

As you go along adding parts to the schematic, SAVE FREQUENTLY. Eagle likes to crash especially if the computer is older or has been on for a few hours and it would be frustrating to lose hours of work.

Print this item

Come in where it's warm!
A warm welcome to those interested in tube and hi-fi audio! Fans of Kevin O'Connor's The Ultimate Tone (TUT) book series form a part of our population. Kevin O'Connor is the creator of the popular Power Scaling methodology for amplifiers.
Please remember these three principles: respect, sharing, community.
Not familiar with The Ultimate Tone book series? See discussion topics, or click here to visit London Power/Power Press Publishing.

Forum Hosted by London Power
London Power logo