Vintage John Linsley-Hood ‘1969’ amplifier build

I've only managed to read the first post or two, and then I decided to find an electronics for dummies playlist on youtube, to help me make a little more sense of this.
Can't wait to follow along.

I had a completely unexpected evening of tinkering last night round at a friend's house. He's also an electronics geek and is currently developing his own audio system consisting of custom bluetooth and DAC based around an ESP32, with separate bass/mid/treble amps in active speakers with DSP (digital signal processing) so he can run a flat or custom response without building boxes to the usual Thiele and Small parameters. This is so he can put the drivers in small sealed boxes and just use the DSP to amp up where the bass rolls off at the bottom due to the small box, and to correct the artefacts that are created by using an incorrectly sized box.

Now, the ESP32, custom DAC and bluetooth bit are a bit above my level of understanding, but I do understand the DSP bit. He is simply doing this with a laptop and a basic external sound card with a microphone, and running a bit of software called REW. In very simple terms, the laptop plays a sweep of frequencies from low to high and listens to what comes back over the microphone. It plots the relative sound levels out on a graph and this is used to correct the frequency response, ie the dips are boosted to bring them up and the peaks are attenuated to flatten them back down.

All this is set up on his bench in his shed, a complete mess of wires, power supplies, amplifier parts, laptop and scopes. I only took the amp round so he could have a listen but, naturally, it ended up in the shed with scopes and stuff getting attached...

1 Attachment

The interesting bit is how he figured out how to look at the response of my amp. First we looked at the maximum output by putting an 8 Ohm load on the output, attaching a scope and pushing it until it clipped. We found that despite the bias (voltage at VE in the circuit diagram) being set halfway, it actually clipped the bottom of the waveform about 3 Volts earlier than the top. That means the bias can be moved up a couple of volts.

Then we had a look at the frequency response with a sine wave. Completely flat at the top end up to 20khz with no evidence of roll-off. At the bottom end it starts to roll off at about 50hz (despite it sounding like it has a pretty strong bottom end), which is probably due to the small input-coupling capacitor. If I increased the value from 0.47uF to 1uF, it would probably drop the low-end roll off down towards 20-25hz.

There was a tiny bit of ripple present in the power supply which I could faintly hear on my speakers. We could see this on the scope. We could also see a bit of harmonic ripple in the supply voltage at the same frequency that the amp was driving. So that was not visible at low outputs but became just visible at very high outputs. Very interesting.

Then we switched to square waves to look at the transient response. Essentially, it holds an almost vertical rise and fall on the square wave up to about 100khz. A little bit of ringing noticeable at 20khz. Down at the bottom end it shows the usual voltage slope at the top of the waveform due to having input and output coupling capacitors. This is because the capacitors block DC, and the top of a square wave is essentially DC so the capacitors block it from getting through.

This was where the clever bit began. That soundcard and REW software he used can also be used to compare the signals going in an out of an amplifier. That tells you not only what the frequency response is, but also what the total harmonic distortion is and whether it’s second order, third order etc. First he connected the soundcard output to the soundcard input and ran a frequency sweep. That generated a curve which showed the sound card’s deviation from a perfect flat response. The software used that as a correction factor to flatten the response. Then he connected the sound card output to the amp input, amp output to the soundcard input (across a voltage divider) and repeated the frequency sweep. This showed my amp had a completely flat response throughout the audio spectrum, give or take a half a decibel which was more than likely error in the correction factors etc. Total harmonic distortion 0.046%, so it’s better than the original design brief of 0.1% or better.

I’m really happy with that and totally surprised. I think if there are any tweaks to do it might be to replace the input coupling capacitors with 1uF caps and maybe put some small inductors inline with the power supply lines to filter out the ripple. Move the bias point up a little and I think that’s really all they need in terms of electronics. We did also stick a thermocouple on the heatsinks and found the PSU heatsink sits at around 60°C and the output stage transistor heatsinks sit at around 75°C. I don’t mind the PSU running at that temperature, but I do know older transistor designs are less tolerant of running that hot. So I might look at increasing the cooling since these are very rare transistors. I have one spare, but it is likely I might not be able to replace them like-for-like if I pop them.

Mad scientist stuff. If you chase a flat response that obsessively is there a risk that it creates a really sterile amp, though? Like monitor speakers, sort of thing? I have no idea, by the way. Just a question.

I have no idea either. I'm way out of my depth, I'm just playing at this point!

Well, keep the updates coming. And don't zap yourself.

I was using REW for tuning audio in a small room. I didn't realise you could plug amps directly into it and measure input vs. output. That's pretty neat and something I'm definitely not going to be doing any time soon :D

@Airhead put me onto it over here https://www.lfgss.com/conversations/156217/?offset=12550#comment17198516

Yeah I looked it up earlier and it started ringing bells from when you were doing that.

Figured I should probably make the front panel for this and then I can have it running a bit more safely. If you remember back, I'd carefully sliced the frame off the original front panel so I could bond on a new, flat front panel. I cut a piece of carbon I'd made back in my F1 days. From memory this was IM7 fibre, 2x2T weave, 2040 resin, 42%. Strength in the order of 800MPa, stiffness in the order of 65-70GPa and Tg around 200°C. So general purpose carbon for use on the floor around hotter stuff like the engine and exhaust.

Panel prepped by lightly keying the surface and bonded on with Scotch Weld 9323 which is the defacto structural adhesive used in aerospace and F1. It has about a 40 minute working time and cures in 24 hours.

2 Attachments