advertisement


Improving oscillators

I would choose 100 Hz for testing SOAR circuits, but the idea is right.
If you plot spectrum of most rock music, the highest content is around 100 Hz
 
^^
Noted, thanks. [What about organ music though....?] ;)
Some E14 'pygmy' bulbs and sockets ordered. Hope these will 'shed some light on it' over the week-end. :eek: :D
 
Couldn't get this to work properly with pygmy bulbs. 620R was too low (wouldn't start) and 1k1 was too high (oscillating between rail-to-rail semi-square wave and 1V p-p and wouldn't settle) 900R was about right but amplitude varied by typically +/-2% and as much as 5%. At least I have some spares for the fridge / cooker!
Ordered some E10 screw 28V / 40mA bulbs and holders to try as this is close to the original CM327 bulb. I can only assume that the bigger bulb has too long a thermal time constant as when measuring the resistance with a DVM (at about 1mA) this rose steadily from 360R to 376R over about 30 seconds.
 
Decades ago I used to have a test bench with kit like the low distortion analyser, etc. But these days it seems a lot easier to DIY based on fairly basic computer + digital kit a la this

http://www.audiomisc.co.uk/USB/ScopePlus.html

Not only easier to get, but cheaper, and does more tricks for a range of measurements. If I wanted better results I'd just get a better DAC and/or ADC than the fairly modest ones I've been using to develop the program. No need for a benchful. Just a couple of boxes sat next to a small computer.

Compare it with the old kit (and awful photos of me!) here

http://jcgl.orpheusweb.co.uk/history/armstrong/AudioAside.html

Oh well, at least I had more hair then. :)
 
Thanks Jim,
Both very interesting reads.....
(slightly envious smile)
I use computers for work most days but don't really have a clue about them TBH.
 
I can appreciate the sheer engineering elegance of ideas like using an incandescent lamp to stablilise a low distortion analyser. My old THD kit used to work very well indeed. And these days when I do build a 'real' circuit I tend to just make very simple ones to scratch an itch. But TBH modern decent amps, etc, should give needlessly low THD. The remaining problems tend to be ones like the example shown by the 'wave from hell' or intermodulation, etc, though. Hence the usefulness of more flexible kit. And in practice I fear the real problems tend to be at the ends of the chain:

1) How good is the recording/broadcast? i.e. source material.

2) Speakers and listening acoustic.
 
Jim,
Noted and appreciated. You comments about IMD etc. concur with the instrumentation engineer at work in that IMD is much more noticeable and deleterious than a bit of second / trace of third harmonics. This for me is more an exercise in as-you-say 'scratching and itch' as much as anything and hoping to learn something along the way. I suspect you are spot on WRT the importance of the two points above too. Especially the latter if my own experiences are anything to go by. I have an awkward (square) room that can sound different with just a small positional movement. I also have some clearly good recordings, some average and some poor recordings, although room and gear can affect this. It's also very hard to know, beyond a certain point where to look next in the chain to improve things. Plus we never know what a recording is meant to sound like. As I've made improvements, I am now aware of layering of instruments which used to sound like one maybe slightly fuzzy instrument. A good example of this is Clapton's 461 Ocean Boulevard.
 
TBH I've never lived in any house where the sound from a stereo *didn't* change when I moved my head as far as an inch! Despite the measures I take to tweak the acoustics - e.g. hang a rug on a wall, suspended via curtain rail so it is held away from the wall.

And,, again, one advantage of the approach I've taken recently is the ability to generate 'arbitrary' waveforms. e.g. Use a pair of 'antiphased' impulses or bipulses from L and R and then mic what reaches the listening location. This then helps find the spot that is equidistant from the two speakers.
 
I can concur with that! My hearing is not fantastic nowadays either (begins to roll-off from about 5K and gone by 9-10k). Only thing that this seems to affect is 'shimmer' and similar which I can no longer hear. For example, the 'tinkle' on the intro to Money (Pink Floyd DSOTM) is now nothing more than a sort of 'crackle'.
Much to the wife's disgust, I have lots of 'stuff' at the dead-end (or rather dead corner as I am arranged diagonally) and have a well curtained window one side (wall) and a masonry wall the other. Very un-symmetrical layout but would struggle to fire across due to window / radiator position relative to doorway. The 'stuff' helps with standing waves etc. Only way I could get reasonably symmetrical would be to have gear/speakers on window wall and sit opposite with door to right. But that would make using the curtains near impossible! If I move sideways by about 1 inch I get a prominent peak but if I move the other way a similar distance, I get another prominent peak at a different frequency. This plays particular havoc with human voices which can suddenly jump sideways!
Only reason I wanted to be able to measure distortion was for making small amplifier modifications so I have a bench-mark to compare to before / after, or more likely mod one channel and compare to other back-to-back.
That sounds an interesting technique, to use a sort of 'ping' to measure the 'timing' between speakers. Must admit that my speakers are a bit of a lash-up. Mids mounted recessed flush on the front of the sub-baffle with a small, sealed box behind them. Tweets just beneath mounted on the rear of the sub-baffle (so that the mid / tweet coils align vertically) with the tweeter hole lined with 10mm wadding to try to control diffraction from the cut-out / baffle corner. This then sit atop a JPW AP2 laid on it's side acting for bass duties with the coil of the bass units about 37mm in front of the coils in the mids / tweets so that they all sit on an imaginary radial curve from the ears at the seating position in a vain attempt to time-align the drivers! I have slowly refined the x-over over the last four years. No measuring equipment, just ears / recordings. I'd often wondered about a technique to measure the relative timing of the drivers, but I am particularly dumb with computers!
 
QUAD used to use squarewave nulling to check that an ESL was similar to a reference ESL. And if you have a pair set up in a good symmetric room acoustic when you play antiphased voice/music material you get a very obvious audible result when you are the equidistant point in front of a stereo pair. Sound gets quiet, oddly ghostly, or seems to come from behind/inside your head!

It is very difficult to time align conventional speakers for a number of reasons. The obvious one is that the inherent delay in each unit (tweeter - mid - bass) tends to be very different. Another is that their radiation patterns are different and vary a *lot* with frequency. Another is the dispersions of the crossover networks.

The ESL63 and its 'children' tackle these problems quite neatly.
 
You don't need a computer as such to assess this if you can find some voice/music recordings that are antiphase mono. Or can use a computer sufficient to employ Audacity to create some off some CDs. Either way, listening can be quite revealing.
 
Yes Jim, heard antiphase mono. Sounds diffuse / spread with no clear source position, but didn't find this that useful. Similar with mono; I have a test and set-up CD but generally I just use a splitter and put say left channel from source into both pre. inputs which guarantees the same signal in both speakers, but again haven't found this that useful either. Could be though that I'm not sure how to interpret the results. I get a general central focus with mono that isn't that tight and some sounds can 'pop out' of the bubble. In some cases I have even had one of those sounds where it say jumps out as though by the left ear, despite only playing through the right speaker! These can only be room nodes AFAICT.
I'd read the Dickason 'Speaker cook-book' and how time-alignment issues cause peaks / troughs in the freq. spectrum and that is why I'd chosen the path above, but as you say it isn't just a case of the x-over affecting phase. The inductance of the driver at the bottom end of the range (esp. bass units) 'delays' the signal, which is almost like the driver moving backwards at the bottom of it's range. Maybe that's how you can get away with mounting the tweeter on the front of the baffle with the Voice Coil 25mm (or more) in front of the mid / bass V/C? [25mm is 90 degrees at 3400Hz, thus in the 'region' of a typical x-over]
Would be interested to find out some more about techniques used in speaker set up (e.g. the anti-phase mono you mention). Is there anything published?
Back on the subject of oscillators, I got some 28V 40mA bulbs (or so I thought). These measured about 30 ohms (cold) which surprised me TBH. Turns out that at 28V, these conduct 120mA so unfortunately not as advertised which was a disappointment. Anyway I ended up using a 'night-light' bulb of 7W @ 240V (measures about 500 ohm cold with the DVM). Takes about 30 secs for the amplitude to settle after switch-on but once settled and warmed-up for about 5 mins, it seems to run with the merest of amplitude variation (1%-ish?) at 3.2V RMS and 3kHz. Now need to build a twin-T notch filter.
 
If simply listening to music/speech antiphased doesn't give a clear null or 'behind you' then it tells you the combination of the speaker radiation patterns and room acoustic are messing up the stereo imaging. To assess that you'd need to use something MF impulses. However the advantage I have here is that the ESL63s and their children have very well defined radiation patterns. I wrote something about using tone bursts year ago. It may give a steer. Hang on...
 
Here is what I had in mind above. One of the items linked from
http://www.audiomisc.co.uk/ArchiveMagazine/index.html
is
http://www.audiomisc.co.uk/ArchiveMagazine/22/Testing.html

In that case it uses a series of tonebursts. By recording the source being played into the amp in parallel with using a mic in the listening seat to capture the result it can measure the phase/time-of-flight as well. In this case I used tonebursts, and did it single channel. But by using other waveforms and L/R/L+R/L-R you can characterise the system more completely. In that case I showed you can simply burn test waveforms to a CD and use a CD player as the source. These days I'd use something like my ScopePlus program and a USB ADC+DAC like the 2i2.
 
Measuring distortion down to low levels is either expensive (if using commercial gear) or involves some pretty difficult DIY if going in that direction. Apparently there are these new fangled computer things that can make things much cheaper... not for me though.

Most reasonably priced (as in <£2-300 ish) distortion meters are only useful down to about 0.1% THD or so. I have Marconi Instruments and Dymar commercial units but don't use them due to just this (actually I redesigned part of the Dymar unit so it can now measure down to about 0.02% ish IIRC but still don't use it). My daily "go to" distortion meter is a home made one designed by Ted Rule and which was published as the Practical Wireless "Durley". It's available on line but is not for the feint hearted. This can measure down to about 0.0014%

My LDO is a Wein bridge based unit which is stabilised by a resistive opto coupler (actually an LDR and a jumbo LED I glued into a black plastic tube) and which is driven by op amp circuitry which optimises response and settling times for each frequency range. It has <0.001% THD (possibly another zero on that even as it doesn't register at all on the distortion meter!)

I believe Black Star now do an LDO at a reasonable price.
 
Interesting stuff Jez. The Jim Williams article talks about using an LDR/LED and associated components for ultra-low distortion instead of the bulb.
Anyway, eventually got mine working. -80dB residual just achievable, but it drifts quickly and before you know it, you have only -70dB and getting worse. I'm going to get some low ppm resistors for both the oscillator and the twin-t notch filter, as it is clearly drifting. With it set reasonably well, at first switch-on and given a minute for the amplitude to stabilise, it shows about 60mV of the fundamental which then slowly diminishes over about 15 mins or so until you end up with what looks like mostly 3rd harmonic (looks like a string of M's on the scope at 3x the fundamental) of just under 1mV p-p, then you start to see the fundamental start to return. More work required!
 
Well after much fiddling I'm still scratching my head....
The twin-T notch uses feedback to the bottom of the tee to control notch-depth and I found that reducing the level of feedback made it less twitchy. I also found that because I'd used well-matched capacitors that I could put some shunt resistance across the variable resistors (for tuning) that made them less sensitive and could sustain about -80dB but for only about 30 secs before this initially worsened and then the fundamental started to reappear (about -70dB). At first switch-on, I found I would have to wait for about 20 mins for it all to stabilise and then adjust the notch to get it 'on song' but then have to go back from whence I'd come to keep it on song. Blowing lightly on the oscillator circuit and then the notch circuit shows that they go in different directions with temperature. The oscillator seems to go up in frequency with rising temperature whereas the notch seems to go the other way so it is not self-compensating, just the opposite. Anyway, seems the trimmers I'm using are now starting to play up as if you try to tweak it, the signal will jump as though you have gone a 1/4 turn. As it is I have about 7 ohms per turn on the trimmers with the added shunt resistance. 1/4 turn (<2 ohm) is significant with 15k on the side legs and 7k3 on the mid leg of the notch filter! Don't really want to spend circa £10 each on a pair of proper pots, but even Bourns say their trimmers are only good for 200 cycles.
 
I'd take a different approach if I were you... A Wein bridge stabilised by an RA53 thermistor will give better results than the bulb, probably good enough in fact, and you can clean it up further if you wish to by simply using low pass filters after the oscillator. Modern op amps can have super low THD so don't be worried about adding distortion from these unless you are looking for <0.0001% performance. Use something like voltage controlled voltage source filters rather than Sallen and Key to avoid common mode distortion from the op amps.
Also look up LDO designs from the likes of John Lindsey Hood, Roger Rosen and Ian Hickman which you should find in old WW issues available online.

On the measurement side it's normal to have to chase the null when trying to measure very low THD.... possibly why so many commercial THD meters only go down to 0.1% FSD as their most sensitive scale. I have to continuously readjust the null to measure down to 0.002% levels with my own meter. It will only stay nulled for maybe 5 seconds but it's enough to get a reading.... mind you the 6 pots that null the bridge (yes 6!) should have been replaced a while back... waiting for some more of those tuits to arrive! It's no problem on the 0.1% FSD range but when switched to 0.01% FSD everything is obviously ten times more critical! If it wasn't for pesky customers wasting all my time by sending stuff for repair and allowing me to earn a living I'd have more time to do proper electronics and would love to design an auto nulling THD meter!

With a good Wein bridge oscillator you should be able to get it so once set to say 1KHz and after say 30 minutes warm up it stays at 1000Hz +/- something like 0.1Hz all day long.
 
Thanks Jez,
I now realise that my choice of bulb was far from optimal and could have been much better.
However, think I'll give the JLH design (Hood_Wien-Bridge_Oscillator_With_Low_Harmonic_Distortion_Wireless_World_May_1981.pdf (proaudiodesignforum.com)) using two op-amp stages and an NTC thermistor a whirl as I can adapt what I currently have (based around an AD712 twin op-amp, JFET inputs) although the details around the thermistor and associated resistors are sketchy. Bought some 2k7 (at room temperature) to play with.
 
You must log in or register to reply here.


advertisement


Back
Top