Distortion meter project

[retailer]

Another small update - today I removed the onboard 5534 IC that makes up the oscillator and ran through the null procedure with an external oscillator - starting with a 1V reference I was able to tune the notch filter to a steady meter reading of 3 with the attenuator set to it's lowest setting - 1mV, and with VERY careful adjustment I could get the meter reading under the 1 mark for 5 or 6 seconds - the fine and null controls are very touchy, I found the best way was to switch the attenuator back one step to a less sensitive range - adjust the FINE for minimum meter reading and then switch the meter back and adjust the NULL, so it looks like getting a distortion reading of 0.01% or lower is possible once I get the on board oscillator and notch filter frequencies matched.
In choosing the capacitors I used a universal component tester and probably this is not the best thing to use when testing components for tight tolerances, I found testing and re-testing the same capacitor gives results that vary by a few percent each time which is not much but in an instrument like this it probably makes a difference between being able to discern 0.01% rather than 0.04% - I haven't tried this but perhaps testing the same capacitor 4 or 5 times and averaging the results might work - in any case I'm going to go through the capacitor selection process again but this time I'll use my TF2700 bridge, it does take a lot longer but I know it gives repeatable results.

[retailer]

Another update on the distortion meter, rather than follow the info in the magazine article I spent a good few hours with the RLC bridge and was able to sort through the capacitors (I had 20 of each value) to find 4 of each value for each of the 3 spot frequencies that were within 1% of each other, some were already installed on the board so out of the 12 I had to replace 5. Before the board was replaced I checked each of the spot frequencies - 101Hz, 1.01Khz and 10.12KHz, with the Null and Fine controls centered I checked each of the filter spot frequencies to make sure I wouldn't run out of adjustment, and found it was better but the Fine control was still almost fully clockwise, would I have enough adjustment ?
Going back and checking the circuit I could see the 10K Fine pot had 2 resistors attached a 68K and a 22K this changed the taper so it would mimic a log taper and also the overall range of the control was no longer 0-10K but rather 6.2K to 8.7K a spread of 2.5K, most of the 2.5K was in the second half of travel when turning clockwise.
The filter cut off frequency is set by the switched capacitors and the 10K Fine control in combination with R9(27K) - doing the maths the Fine control and R9 should add up to approx 33.2K, R9 is 27K the Fine control and it's resistors should therefore be approx 6.2K and still have some adjustment in hand - my multimeter showed it was 6.2K when full clockwise - when centered it was around 8k. Simple I thought just reduce R9 (and also R10) to 25K, the Fine control when centered plus R9(27K) would add up to 33K which would leave adjustment room for tuning the filter. A 330K was soldered across R9 (and R10) followed by a trial - yes I had plenty of adjustment and it all looked good until I started step the attenuator down towards 1mV, I found it hard to settle the meter needle, it felt much touchier than before. I was a bit puzzled at first and it took me a while to work it out, with the value of R9 now being lower the Fine control now made up a slightly bigger portion of the overall 33.2K resistance I was aiming for so while I now had some adjustment I had also made the Fine control more touchy - gain one thing but lose another, what I should have done was alter the values of R7 and R8, the two resistors attached to the Fine control pot - now that the capacitors are close tolerance I may not need a 2.5K spread for tuning, possibly less then 2k, I'll remove the 330K resistors piggybacked onto R9 and R10, and play around with the 2 resistors soldered onto the Fine control R7 and R8 - if I make R7 27K and R8 27K it will give me a spread of 1.5K with 33.2k coming in at about the 75% mark - more work for tomorrow.

[retailer]

Yet another update on the distortion meter project, I checked the resistance range of the Fine control at all three spot frequencies and found it's resistance varied by only 700 ohms to be able to achieve a good null, to allow for slight drift I added 500 ohms to this and decided that the Fine pot only needs to vary between 5.5K and 6.7K to be able to tune the filter, rather than sit with a calculator and work out values for R7 and R8 I simulated it with Electronic work Bench and settled on 22K for R7 and 33K for R8. The total resistance of R9(27K) plus the Fine control pot with its resistors now varied between 32.7K and 33.9K - my target value was 33.2K, as the Fine control pot is dual gang, R11 and R12 also need to be changed. With the new resistors installed I found it quite a bit easier to null the meter - it was still touchy but now much less than it had been. I also decided to replace the 0.5K Null pot with a 10 turn 1K unit that's been lying around in my used potentiometer shoe box for a few years, it's a bit scratchy if you hit the wrong spot but overall it is a worthwhile improvement and will do until a new one arrives.

I reasonably happy with how the notch filter performs now, nulling the meter to 0.01% is repeatable whereas before 0.04% was repeatable but 0.01% was hit and miss, the oscillator however is a bit of a puzzle, according to the meter the distortion level of the oscillator is around 0.05% - whereas my home made Wienbridge oscillator, which was a construction project from Australian Electronics magazine has a distortion level of better than 0.01%, the limit of the distortion meter, this uses a 'proper' glass bead thermistor against the distortion meter's 12V 50mA globe, not sure why it is like this as the magazine article states they achieved around 0.003% distortion, something else to look into.

Of course I couldn't wait for the new Null pot to arrive before giving the meter a real world trial so I did a quick 1KHz test on one channel of a Dynaco SCA35 amp, using the external oscillator as the signal source - at 1 watt at 10 watts measured distortion was 1watt 0.095% (just below 0.1%) and 10 watts 0.11%. I still have a bit to do but once the new null pot is in place I plan to do a comparison between a Leak TL12, which can be regarded as the benchmark, the Dynaco and the Jason J2 10 - all three are powered by PP EL84's

[Diabolical Artificer]

Nice work. I've been following the thread but have had nothing to add. 0.095% THD @ 1w for a valve amp is pretty good going as at that low level LF 50/100hz artifacts contribute to most of the noise/hum & effect the reading I've found.

Andy.

[retailer]

I spent a bit of time with the distortion meter testing the Dynaco and a small Japanese SE amp I have, mostly at 1KHz but when I switched the oscillator to 10KHz I could see on the scope screen the frequency had drifted, it started out at 10.12 KHz and after 15 or 20 min of continuous operation this had risen to 10.4KHz, just outside the range of the null controls, the 1KHz and 100Hz remained at 101.2Hz and 1018Hz - looks like the 10KHz 470p blue ceramic capacitors will have to be replaced, with a limited range of adjustment in the filter the oscillator needs to be very stable. I also noticed that the oscillator takes 5 min or so for the distortion level to stabilise, at first turn on at 1KHz it is up close to 0.1% gradually dropping down to around 0.05% over the course of 5 or so minutes, I may have a bit of a play with RV7 which is in the feedback path of the light bulb and see if the amplitude affects the distortion level. While searching through my used potentiometer shoe box I came across a vernier drive, seems to be a round 4 or 5 to 1 ratio, I mocked up a simple bracket out of cardboard and it looks to be a reasonably easy fit into the meter case should I need to do this at a later date,

I didn't really know what to expect from the Dynaco - it doesn't seem too shabby - it is more or less a brand new amp, I purchased it as an unbuilt kit and apart from some testing it hasn't had any use since I completed it some months ago.

[Diabolical Artificer]

What light bulb are you using? I found the best results with a Wein bridge oscillator were 28v automotive bulbs. The 1khz ish Wein bridge oscillators I built with TL074 op amps remained pretty much bang on frequency wise but I did find they have a sweet spot (amplitude) as regards very low THD.

I even used an ECC82's heater as the lamp/auto gain element though this isn't practical in application.

Andy

[Radio Wrangler]

You can scale the impedance level of the arm of the bridge with the light bulb in it, and the operating voltage level to suit available bulbs if you need to (Just not LED and compact fluorescents )

David

[retailer]

I'm using the bulb recommended in the project writeup 12V 50mA - I checked and while not bang on it is 47mA@12V, I played around with RV7 amplitude control and found a sweet spot where the distortion measures at 0.045%, it should be better than this, it's a long way off my home made weinbridge oscillator which uses a 'proper' glass bead thermistor. What is notable is that if I set the amplitude to 2V as per the project setup instructions the oscillator bounces and won't stop until I increase the amplitude to approx 3.5V - the sweet spot is around 5.5V. I've got a bunch of output transformers to wind, once they are done I'll build the circuit on a bread board to play around with it. I'm also going to bread board the attached circuit and see how it performs

[Radio Wrangler]

That's actually not a Wien bridge. Notice that there are two concatenated, buffered series R, shunt C sections.

The Wien bridge is just that. A bridge amplifying the difference between a resistive potentiometric branch and a branch incorporating series RC and parallel RC sections. In the Wien bridge, the amplifier runs at high gain and a single feedback path drives both the resistive and RC branches of the bridge.

There are similarities, but there are differences in the analysis. It doesn't all come out in the wash.

David

[Jez1234]

Appears to be series connected all pass filters to achieve Barkhausen criteria. I'd stick to Wien or Twin T oscillators for this personally (although good results can be obtained using a state variable filter based oscillator). There are a variety of issues with that design above which would concern me; rails rather too low at only +/- 5v, lack of isolation between op amp and rectifier and impedance's too low for TL07x series op amps to drive optimally.

If the OP checks out old Wireless World and Electronics World issues on worldradiohistory.com there are several really good designs from Ian Hickman (also wrote as Ian March) and John Linsley Hood. If wanting to go the all pass filter based route then there is a very interesting design from Roger Rosens which out-phases the 3rd harmonic to get very low distortion here: https://worldradiohistory.com/UK/Wir...ld-1982-02.pdf

[Radio Wrangler]

The designers of the HP8903 went for a state-variable oscillator (and a matching one as the notch filter for the distortion measurement side). They said this was because of a small advantage in distortion.

They're all R-C oscillators and from the point of view of phase noise analysis have a figure of merit Q of less than one. The HP one is tuned using precision analogue multipliers under microprocessor control which tunes the frequency and tweaks the amplitude. They can also get the notch a little narrower this way which helps accuracy on SINAD.

David

[Jez1234]

I spent an afternoon looking into the possibilities of making a self adjusting THD meter using four quadrant multipliers a few years ago. It was to use an "analogue computer" ie op amps to control it. Unfortunately IC's for doing things the proper way (ie analogue not digital!) are becoming scarce in some fields and analogue multipliers seems to be one of them. Complexity, price and difficulty in obtaining the multipliers, and having bigger fish to fry meant it got no further. Switched capacitor filters looked like they may be worth investigating further I seem to remember...

[retailer]

Another update on the distortion meter project - my 10 turn pot arrived and was fitted and I decided to utilise the vernier drive I had on hand to make the Fine adjustment a little easier - as mentioned in previous posts testing the internal oscillator with the meter shows it may not be performing as intended by the magazine project authors, distortion seems to be on the high side, to see if I could improve this I decided to swap out the Texas Instruments branded 5534 IC for some other IC's I had on hand firstly a Philips branded 5534, an OP07 and lastly a TL071, the TL071 is the only one that has previously been used in other projects - the Philips 5534 failed to produce any oscillation - adjusting RV7 made no difference, the OP07 was the same and the TL071 came up with a square wave which couldn't be adjusted. I have around 10 of the 5534's all TI branded plus the one Philips - I thought if the Philips IC failed to work at all then possibly I might get better performance by trying all of the TI 5534's one by one, so I started swapping them out and on the 4th swap it was noticeably better - I let the oscillator 'warm up' for 5 mins after which I found I could null out down very almost to .01% much better than the .04% or so that had I been getting - I've had the 5534's for at least 15yrs possibly longer, all purchased at the same time most probably from Jaycar a local hobby electronics store, I don't think I was buying stuff on line back then. I'm probably going to call it quits on trying to get better performance, it's taken a lot more of my time than I had envisaged it would, my main interest is in valve amps and as the instrument stands it is plenty good enough to measure down to 0.1% or so that is typical of better built valve amps.
All I need to do now is tidy up and engrave the front panel and fit a handle and the tilting bail type feet that I have on hand.

[Radio Wrangler]

Just an odd thought about the frequency drift. "Blue ceramic 470pF capacitor" ?

There are several different ceramic formulations available, usually known by their temperature coefficient code. NP0 or C0G grade have low temperature coefficient. X7R and the others are A) cheaper and B) pack in a lot more microfarads/volts into the same size package but hey have orders of worse drift over temperature. some of the, are also fiercely microphonic.. in fact some meterials are USED as acoustic transducers.

For oscillators, filters, tuned circuits, it's best to stick to NP0 or C0G and stick to a reliable source who will not substitute the cheaper type. Use the other sorts for decoupling in non-critical places.

David

[trobbins]

Well done for battling through. There is obviously a lot more to vacillate over when trying to eek out more from the wein bridge notch and oscillator. The mid-1960's AWA distortion meter (based on Wein bridge notch) I restored earlier this year uses 4-20pF trimmers, and needed iterative trimming of all variables to get the notch portion down below -65dB, and the later model AWA distortion meter was similarly quite sensitive with a titch of frequency drift. Stabilizing local temperature, and especially around the bridge caps, after seeing if you have any better tempco caps, maybe a worthwhile last tweak to avoid drifting results (even if DUT distortion is 10x the best floor level you can initially set up).

[retailer]

In retrospect it was probably a mistake to buy 'ordinary' 5% or 10% caps in lots of 20 and then go through them to find close matched sets - for each of the spot frequencies 2 caps are needed in the oscillator and 2 in the filter, the filter and oscillator frequencies should match exactly, or as near as you can get to this - the greater the spread between the filter and oscillator the greater the tuning range needs to be on the filter, this makes it harder to tune using ordinary off the shelf potentiometers.
1% through the hole pcb capacitors don't seem to be that common but I did track some down at Mouser, the smaller values are quite reasonably priced at around $2 or so each only the larger values like a 100n capacitor is a bit pricey at around $8 ea, but overall it may only be an extra $10-$15 to get the 1% caps rather than the 60 capacitors if you buy 20 of each value.
As the 47pF caps need to be replaced I may still replace the whole lot with 1% caps at the time I fit the front panel once it is engraved.