Distortion meter project

[retailer]

For those that are interested I have a small update on the distortion meter project, the board populating went easily - the article recommended and described a way of selecting matched values of capacitor for the oscillator and filter - to do this I ordered 20 of each value, measured them and then paired them and managed better than 1% matched.
With the metalwork/hardware finished and boards all populated it was time to wire up the various switches/pots etc, it did take some time as the enclosure I was using wasn't very deep so it didn't leave a lot of room for access to the row of pcb pins along the front edge of the board, once wired up I decided to test it in stages starting with the oscillator. I powered up with no ic's plugged in, both the power supply rails on all of the ic's were close enough to + & - 12V so time to try the oscillator, result was zero output on the scope across the entire range of RV7 trim pot, double checked the switch wiring it all looked good, trying all switch settings I managed to get a signal of sorts that looked nothing like a sinewave, the oscillator ic was starting to feel warm to the touch so something wasn't right, I could still test the mV though.
With the oscillator ic removed I powered up again and connected my signal gen to the voltmeter input, that at least appeared to be working - even though the scale was still not installed I tried to roughly set the FSD - no joy here, so I had some error checking to do on the mV meter, I then decided not to go any further until these two issues were sorted.
The oscillator issue was tracked down to an error on my part with the pcb connector pins, I had missed drilling a hole in one place and drilled an extra hole in another resulting in all of the oscillator switch connections being displaced by one pin, with this fixed and the oscillator working I went through the setup procedure which says to set RV7 trim pot to give a maximum output of 2V, however I couldn't get 10KHz to work reliably at 2V - if I powered up with it set to 2V it was fine but switch to 1KHz and then back to 10KHz it would bounce in and out of oscillation so I ended up setting it to 3.5V which produced a nice stable sinewave, and bringing the output back to 2V by wiring a voltage divider across the output socket. Once set up, the frequencies were all quite close to the design values, 10.02KHz, 1.01Khz and 102Hz, I could now move on to the mV meter.
The article describes the mV meter circuit - 1mV input should give 256mV output at pin 6 of IC4 a gain of 256 - I had 198mV - the gain of 256 was determined by R40 and R41 both were spot on so it was a bit of a mystery, I then noticed I has swapped IC3(TL071) and IC4(5534), I couldn't see that this would make a difference, or at least it shouldn't, but after swapping them over into the correct positions I now had my 256mV (268mV to be exact) and I could now roughly set the FSD, this can be completed once the scale is in place.
I still have the test and setup of the filter- print and fit the meter scale - tidy and make a loom of the internal wiring, some of it is too long, as well as a small amount of work on the case to complete ( feet handle etc) and a set of knobs to source - total cost so far without knobs is approx $55, the knobs look like they will double the total.

[ian_rodger]

I've often thought that a stand alone distorion meter would be of use to me. As pointed out earlier in this thread, the magazines are available to download. I've tidied up the PCB outlines and etched a set of boards using toner transfer in preparation to build this as a winter project. I can post copies of the artwork here as jpegs if anyone is interested.

Also have one spare copy of "press'n'peel" transfers I can make available for the cost of postage.

[retailer]

Another update on the distortion meter project - completed the meter scale and all of the wiring and setup, so time to see how it operates, I knew the mV meter and oscillator were now operating ok so I jumpered the oscillator output to the meter input for a first try at checking the distortion of the oscillator at 100Hz. In practice with the function switch set on THD the Fine and Null controls are adjusted for minimum meter reading - switch down a range and again use the Fine and Null for a minimum meter reading. I found the Fine control was at it's fully clockwise position so I only had the Null control to adjust, the minimum meter reading I could achieve was 30mV, or 3% distortion not a great result (the reference or start voltage is 1V). I tried 1KHz and 10KHz and while it was marginally better I still found the Fine control was set fully clockwise.
Next I connected my bench signal generator at 100Hz - with the Fine and Null controls set to mid position I adjusted the bench signal generator frequency for minimum meter reading, to match the signal source and filter frequencies, continuing on with the procedure I was able to get down to 12mV or 1.2% distortion. My bench signal generator uses a function generator chip, not sure which one, maybe 1% or so distortion is right. Next I tried my home made audio signal generator at 100Hz, using the same procedure to bring the oscillator and filter into line with each other, I made this oscillator many years ago it is a Wein Bridge type that uses the R54 thermistor, the minimum meter reading was 1.9mV or 0.19% distortion, I may be wrong but I believe these types of oscillators have the most distortion at lower frequencies so 0.19% may be right, I couldn't check at higher frequencies as the frequency select switch had seized probably through not being used.
It's clear I'm going to have to adjust some of the components to bring the oscillator and filter into line with each other so the Fine control is not operating at the extreme clockwise position, quite possibly I may have achieved a better first attempt result with the internal oscillator if there was some adjustment available in the Fine control.

[Smithdoor]

Looks nice

Dave

Originally Posted by retailer

A small update on my distortion meter project - adventures with engraving the front panel - I had planned to have all of the hardware/metalwork finished by now and be starting final assembly but encountered a series of issues with the engraving - broken cutters, engraving too deep, bad choice of engraving font etc. etc. Fusion360 has 2 ways (that I know of) to engrave, one method uses true type fonts (same as Windows) - the text is placed on a 3d drawing of the front panel and then given a negative height this gives it the appearance of being cut into the panel, once all of the text/numbers have been placed in the correct position, the cutter type is chosen, the text is selected and Gcode is generated for use on the mill. Engraving cutters are quite simple - 'D' bits taper ground to a point, I was using 30deg cutter, when engraving true type fonts Fusion tries to maintain the correct width of the font by varying the depth of cut as needed, choosing a fat or bold font will results in greater depth of cut (cutter is tapered). I used this method as I'd used it in the past and had acceptable results.

The first panel made from Al was a fail because of an error on my part in defining the tool path which resulted in some text being engraved as .5mm deep hole, this happened almost straight away, with the first few engraved numbers, I noticed and stopped the job but the damage was done I had to discard it and start again. Had I let the engrave proceed I would have been alerted to other problems, which doomed the second attempt to failure.

Searching for a second piece of suitable aluminium I came across some black acrylic 3mm sheet, just enough for a panel plus one extra in-case I stuff it up. The first piece of acrylic was a fail - my choice of font was bad - I chose a compressed bold font which resulted in engraving that went far too deep, Fusion plunged the cutter in to maintain the width of the font, as the font was compressed the letters were too close together and ran into each other, the fine pieces of acrylic between the letters just crumbled - I could read the numbers and letters but only with a bit of imagination - this is what I would have noticed if I had let the first attempt continue - I still had one other piece of acrylic left, perhaps - I should view some youtube videos on how others have engraved.

One youtube video also described the second method of engraving, in Fusion this is called trace - this uses single line text - it has no width - the text width is set by a combination of the cutter taper/angle and also the cutter's depth of cut, this is more like traditional pantograph engraving where the cutter depth remains constant, knowing the cutter angle one just sets the depth of cut to achieve the correct width text/numbers, the choices of font are much smaller but this is probably more applicable for panel engraving, I decided to try this but was faced with the prospect of going through my design and edit all of the numbers and text from true type to a single line font, rather than miss something I chose to redesign the entire front panel. With the new setup/gcode I had the depth set at 0.3mm this would allow for slight upward bowing of the acrylic panel - with the cutter being quite short, the mill chuck is very close to the panel and this obscured my view, so I stopped it after a 1/2 a minute or so to check - it looked great, but something wasn't right, the text was placed too high on the panel and didn't line up with the position of the rotary switch - I won't say what I was thinking - back at the computer in front of fusion360 I could see I my origin point was 10mm out - I'll complete all of the engraving, on top of the last error so at least I'll see if there were any more issues, for a while I entertained the thought that maybe I could fill the error with black paint or epoxy but acrylic sheet is not that costly so for now I'll use this panel and when my new piece arrives I'll redo it.

I know some are thinking surely just printing the front panel artwork on paper and gluing it on would be easier - yes I agree but then I would have missed out on all that fun, forewarned is forearmed so for the next project, if there is one, I'll know what not to do.

[Smithdoor]

I am on welding site where looking at the distortion of inverters. Then they could not find the old fashion generators distortion data.
Simple answer the old fashion is what resting against. It is a sine wave
They can not wrap there head around that part.
I gave up.

Dave

[retailer]

If any forum members are thinking of building this project then you may want to give some thought to the type of switch used for the meter attenuator, I used a low cost unit from Jaycar, 1pole 11position, and have found the meter gives a good fsd 'kick' each time the range is switched up ie less sensitive, my switch has break before make contacts, possibly a switch with make before break contacts may help prevent the fsd 'kick', I have one on order and intend to try it.

[retailer]

My THD meter is sort of finished but I'm not really satisfied with the performance - there are times when the meter needle oscillates at around 1-2Hz the amplitude of the oscillation is very small in the region of 70-80 uV, I've tried but can't really see it on a scope and running the unit from a pair of 9V batteries makes no difference, without access to better test gear it's hard to tell where and why this comes from, so not being 100% satisfied with the project I've been investigating alternative oscillator and notch filters. I came across an alternative oscillator circuit from Amateur Radio magazine March 2010 and built up a breadboard example - still a Wien Bridge but uses a LDR/LED opto coupler to stabilise, an internet search brings up quite a few similar circuits using the opto coupler idea, I also investigated an active twin T notch circuit, there's a good article on a twin T active filter at http://www.tronola.com/moorepage/Twin-T.html, it's very similar to the one at https://www.sound-au.com/project52.htm. I built up a breadboard example set at approx 1Khz, early testing of the breadboard version with jumper wires all over the place, looks like this is a better proposition so I'm going to build one up at the 3 fixed frequencies.

[trobbins]

It may be worthwhile quickly using an independent spectrum analyser based on your PC and soundcard and software like REW, which I've used to restore vintage Wien bridge oscillators, and Wien and Twin-T notch filters. Luckily the local AWA made very sturdy and low noise notch filters within their vintage noise and distortion analyser equipment from the 1950-70's, which are nowadays considered as boat-anchors to be tossed in the bin, but hide excellent tunable notch filters (its just the surrounding active electronics that nowadays limit their performance).

Although the use of modern soundcard/software could be seen as defeatist, I find that it helps uncover the original technical capabilities of vintage equipment, as well as any restoration or diy quibbles that may be difficult to discern.

[Jez1234]

If looking for a really good low distortion oscillator for THD measurement may I recommend a modified Heathkit IG18, or to do as I did and build one from scratch to the modified spec.

It is a good oscillator to start with but then Reg Williamson published an article called "The greening of the Heathkit IG18", which was followed a few years later by an article in Audio Amateur called (IIRC) "Gilding the IG18" and which took it a stage further.

I only had the AA article to go on but as part of the extensive modifications it gave the original schematic. It ends up as an LED-optoisolator controlled oscillator using an NE5534 as the oscillator and with a TL074 quad op amp doing the rest of the work including precision rectifier, control loop compensation and LED driver.
The finished thing covers around 15 Hz to 30KHz in four ranges, has switched attenuator and variable output levels and manages around 0.0001% THD.

I've used mine for all distortion measuring since I built it back around 25 years ago now. Yes in conjunction with the PW Durley!

[trobbins]

One advantage of modern day software is that even my 15 year old soundcard can be configured to achieve below 0.0001% THD - the software trick being that subtle levels and phases of harmonics can be added to the generated fundamental such that the resulting generated test tone is 'harmonic distortionless'.

Just saying that this path may be already available and in place to retailer if he has a soundcard with his PC. And to support the benefit of having an alternate software-based spectrum analyser, I note that the IG18 update article used ARTA to verify performance improvements.

[Radio Wrangler]

The softwate method of adding deliberate harmonic content to cancel that already there relies on having a really good distortion indicator. If the measuring side isn't so good, then you wind up setting the distortion of your source to be the complement of the distortion of your measurer. Connect the two together and they show low distortion. Use this to measure the distortion of something and it's not quite the same as having a low distortion source. There will be errors, though they can be small under favourable circumstances, but you need to know about this vulnerability to work out whether it's tolerable. When you push measurements towards extremes, all sorts of limitations come out of the woodwork. All measurements have errors, the art lies in knowing their bounds.

Wien bridges, don't themselves suffer worsening distortion at low frequencies, but wien bridge oscillators as built usually do. These oscillators are not naturally stable in amplitude so some deliberate stabilising mechanism has to be engineered in. The original method was to use a thermistor to adjust the gain in a level-dependent way. Bill Hewlett invented this and used the temperature/resistance characteristic of a common filament light bulb. Later implementations used rectifiers and FETs as voltage-controlled resistors. Some arrangements use 2-quadrant analogue multipliers. Common to all these is a time-constant in the slew rate of the gain control function. At lower frequencies, the gain control starts to track the waveform of the output sinewave, and thus compresses it a little. The effect gets worse ar lower frequencies when the gain control tracks more, compresses more. The solution is to slug the speed of the gain control. This, however makes the amplitude slow to settle after a frequency change. The dreaded Wein bridge 'bounce'. So there is a compromise between settling speed and distortion at the low frequency end.

This is a general problem with AGC-controlled oscillators and one of the reasons why Ulrich Rohde argued against the use of the AGC diode trick in JFET low phase noise RF oscillators, The AGC system bounce exaggerating the phase noise at lower offsets from the carrier. Remembering that the noise creation mechanism also makes amplitude noise as well as phase noise and each can turn into the other.

David

[trobbins]

Some enjoy making an instrument, and never worry about the actual accuracy of the measurements made by the instrument, and in some cases the accuracy isn't the benefit as it is the relative comparison provided by the instrument that is a key benefit. But given time, some take the journey to better appreciate the absolute accuracy of their instruments, and a psuedo-metrology quest starts.

I lucked on a cheap but good EMU 0404 USB soundcard a decade ago, and even got Ed D to bring another one over to have as a backup. Some distortion testing in 2019 (https://www.diyaudio.com/community/t...341272/page-3) used a technique that many start with to determine at what signal levels the DAC or the ADC of a soundcard start to show rising distortion levels. But early last year I too went on a quest to try and determine how low the intrinsic distortions were in that soundcard, spurred on by advances in the REW software, and helped out by the twin-T notch in a vintage commercial distortion meter that many think is a boat anchor. https://www.diyaudio.com/community/t...rtions.381839/

This may seem over the top, but tools like this can come in quite handy when improving on existing equipment An upcoming project I have is to reconfigure a circa 1964 AWA Distortion and Noise meter that is based on a Wien bridge with EF86 surrounding circuitry which I will try and sideline with 5534A opamps.

[Jez1234]

When I built the IG18 based oscillator I had no access to the recommended optoisolator so made my own with an ORP12 and very large LED in a sealed tube. This had somewhat different characteristics of LED current Vs LDR resistance etc compared to the original and I had to change the compensation rather a lot to get stability across all the ranges. It's not straightforward compensation and IIRC has multiple poles and lag lead used (i haven't googled it and article is not to hand). Many years later the LDR started to go dodgy and I managed to get some Perkins Elmer LED-LDR optoisolators similar to that in the original schematics. Of course this then meant going through all the compensation again! I strongly suspect that the original values for this in the schematics are not optimum.
It has fast and slow settling speeds available with much longer integration time used for very low frequencies in order to alleviate the LF distortion issue David mentions above. This makes compensation even more tricky as you can have it OK without the slow integration but it takes off with it.

Whilst it is a superb LDO when working right I may have put the OP off the idea there!

[Jez1234]

Originally Posted by trobbins

Some enjoy making an instrument, and never worry about the actual accuracy of the measurements made by the instrument, and in some cases the accuracy isn't the benefit as it is the relative comparison provided by the instrument that is a key benefit. But given time, some take the journey to better appreciate the absolute accuracy of their instruments, and a psuedo-metrology quest starts.

I lucked on a cheap but good EMU 0404 USB soundcard a decade ago, and even got Ed D to bring another one over to have as a backup. Some distortion testing in 2019 (https://www.diyaudio.com/community/t...341272/page-3) used a technique that many start with to determine at what signal levels the DAC or the ADC of a soundcard start to show rising distortion levels. But early last year I too went on a quest to try and determine how low the intrinsic distortions were in that soundcard, spurred on by advances in the REW software, and helped out by the twin-T notch in a vintage commercial distortion meter that many think is a boat anchor. https://www.diyaudio.com/community/t...rtions.381839/

This may seem over the top, but tools like this can come in quite handy when improving on existing equipment An upcoming project I have is to reconfigure a circa 1964 AWA Distortion and Noise meter that is based on a Wien bridge with EF86 surrounding circuitry which I will try and sideline with 5534A opamps.

I get such urges as well! When I see old test gear and think "if only they'ed thought to put a current source/whatever just there the accuracy/range could have been an order of magnitude better". I nearly got drawn in to doing this with the Advance H1B I recently sorted out. The THD of just the Wein osc itself measured 0.2%. It's thermistor stabilised. It would be very easy to get it to something like 0.01% I reckon and possibly below that.

I have an old Dymar distortion meter which has a most sensitive range of 0.1% FSD but the intrinsic THD of the input amplifier of the unit measured something like 0.05%. It was well out even measuring 0.1% and the lower reaches of this range were useless. I replaced the original input amp with one made around an NE5534 and it was then pretty accurate down to 0.01% Result! Unfortunately after all that I haven't used it in years as the PW Durley is way better still. It's there as a backup though. My Marconi 2331A has similar limitations but I've left it alone....

[Radio Wrangler]

CdS photoresistors don't have a single linear model as far as their controllability goes. The darker their illumination and the higher their resistance, the slower their response gets. So if you're stuck with whatever photoconductor is available, and you don't like the time constant, you can always re-scale your circuit to run them at a different mean resistance. Lower-Z circuits mean lower R and means brighter-lit, larger signal and faster response.

Messing around with ideas for a super-low intermod HF receiver, I wound up buying a load of ORP12 for AGC controlled gain IF stages. The things are nicely linear. I managed to scrounge a few samples of some of the very first 8000 mCd LEDs to light them. They worked well, but I learned a lot about their time-constant shift!

I keep thinking about an audio preamp with photoconductor input (etc) switching and a photoconductor based volume control.

One concern is in their noisiness at the high resistance end of their range. Maybe series-shunt-series topologies and I like active volume controls.... Hmmm

David

[Radio Wrangler]

At a radio club junk auction there can be some college lab disposal stuff turn up and I picked up a pair of basic Wein bridge oscillator and distortion meter. They made a good Christmas present for my brother who at the time was building guitar amps for himself. A distortion meter? To be sure he was getting enough!

I'd already found a nice little Hameg scope for him.

David

[Jez1234]

Originally Posted by Radio Wrangler

CdS photoresistors don't have a single linear model as far as their controllability goes. The darker their illumination and the higher their resistance, the slower their response gets. So if you're stuck with whatever photoconductor is available, and you don't like the time constant, you can always re-scale your circuit to run them at a different mean resistance. Lower-Z circuits mean lower R and means brighter-lit, larger signal and faster response.

Messing around with ideas for a super-low intermod HF receiver, I wound up buying a load of ORP12 for AGC controlled gain IF stages. The things are nicely linear. I managed to scrounge a few samples of some of the very first 8000 mCd LEDs to light them. They worked well, but I learned a lot about their time-constant shift!

I keep thinking about an audio preamp with photoconductor input (etc) switching and a photoconductor based volume control.

One concern is in their noisiness at the high resistance end of their range. Maybe series-shunt-series topologies and I like active volume controls.... Hmmm

David

Interesting about the LDR there. May have helped make my LDO such a pig to compensate!

Using LDR's as volume controls has already been done. There's a model available called the "Stereo Coffee", of all things. LDR's are not that linear though and can generate something like 0.03% distortion due to the voltage coefficient of resistance. Of course some audiophools swear they are like removing a blanket from in front of the speakers

[retailer]

A bit more of an update on the distorion meter, I've identified a few issues - the signal source and notch filter frequencies don't match, so much so that I run out of adjustment on the fine null control, so I need to adjust the signal source frequency so the two match. The other issue is hum and noise being picked up due to lack of any sheilding. The enclosure is a large PVC Clipsal junction box measuring 280mm x 300mm, and the front panel is black perspex, my hand in close proximity to the control pots will change the meter reading. I'm going to line the interior of the enclosure with copper sheet, I have quite a bit of copper sheet which at one time was the flexible connection for a massive spot welder, and I'm also going back to an aluminium front panel.

Originally Posted by trobbins

Luckily the local AWA made very sturdy and low noise notch filters within their vintage noise and distortion analyser equipment from the 1950-70's, which are nowadays considered as boat-anchors to be tossed in the bin but hide excellent tunable notch filters (its just the surrounding active electronics that nowadays limit their performance).
.

Older distortion meters are around and do at times come up for sale, this week I did locate an AWA distortion meter on Gumtree in Williamtown Vic. however the seller will neither post it or agree to me organising a courier to pick it up

[orbanp1]

The antiqueradios.com site had a diy distortion meter a couple of years ago (around 2016), it had auto tuning.
You could look at that circuit for inspiration.
Unfortunately I can not find the posting, but I saved the core article in pdf format that I enclose.

Regards, Peter

[retailer]

Thanks for the PDF it is an interesting read - the auto tuning idea looks to have come from necessity.
I made some headway this afternoon - resurrecting my original front panel made from aluminium and lining the enclosure with copper sheet has made a very noticeable difference - the issue with the built-in signal source and notch filter frequencies being slightly off still remains so I'm unable to do any testing on the internal signal source so I tried my home made sinewave gen and adjusted the frequency to match the notch filter, this sinewave gen should have a distortion level under .01% - I was able to get a 'very touchy' null at .04% - OK, but not great. After a bit of experimenting, I came to the conclusion the internal signal source was bleeding through and affecting the meter reading.
With the function switch set on volts and the input shorted the meter reads zero right through all ranges of the attenuator down to 0.1mV, but with the function switch set to THD and the input shorted I was getting a meter reading which varied from 50 or 60 microvolts up to 0.2mV (3mV range) depending on the frequency setting, as the signal source and notch filter are on the same board the only way to disable the signal source is the remove the IC, which is what I'll do tomorrow and once again check the notch filter with my home made signal gen.