Low-distortion Audio-range Oscillator

Member
Joined 2004
Paid Member
Maybe a cap across R51 or R100 would address the frequency issue , compensating for something. Does cascoding work for FET's used as VCRs? It would open up more ways to control and linearize things if it would work.

I tried a crude simulation and it does seem to work but I'm not sure how well and I don't trust simulations for that.
 
Last edited:
Disabled Account
Joined 2012
@Richard Marsh -- my experience is that two multi-turn trimmers are not needed for R50 and R51 -- leaving one of them as-is and reducing the value of the other by a few hundred ohms and then putting one trimmer in series in order to swing the balance ratio of the two resistors above and below exact match will accomplish the same goals. I agree on the trimmer for R30 -- good idea.

The balance between R50 and R51 does not change the AGC input level at the JFET gate, they are not independent; rather their ratio only affects the feedback 2nd H cancellation. I found that optimizing the 2nd H reduction at 1kHz tended to make the 2nd at 10kHz worse and vice-versa, so selecting a value that optimizes both and living with the compromised result is to me a good strategy for the HP 339A.


Yes, you can do it that way, too. It depends how far down you want to get the harmonics. The 3H is weak and only shows up at >10Khz. It can be further reduced but using a lower value than 2K. The lower the better - as long as the opamp can drive it OK. So replacing both of the 2K values with 1K trimmer helps. Then adjust for harmonic null.

How low did you get the THD on your 339 and using what for the osc opamp?
I recently decided to get my spare parts 339 operational.... It just needed a reed relay and then the tune up started. On that one I only replaced the R30, R51 with a trimmer. And, then did the opamp/parts upgrade... this 339 has the power push button on-off instead of the toggle switch. The push button unit has better mechanicals on the all the rotary switch shafts etc and thus is a better unit to modify. THD of 2H and 3H at 1KHz is --> -120dB each. At 10KHz they are --> -110 and -134dB. This is not the killer 339 I did earlier but using what I did on that first one to do just those changes which make the most improvement.

Thx-RNMarsh
 
Last edited:
Disabled Account
Joined 2012
Maybe a cap across R51 or R100 would address the frequency issue , compensating for something. Does cascoding work for FET's used as VCRs? It would open up more ways to control and linearize things if it would work.

I tried a crude simulation and it does seem to work but I'm not sure how well and I don't trust simulations for that.


yes, there are some small phase shift going on which cause the distortion null to shift away. It can be tried with a variable C trimmer at the appropriate place while watching the thd meter.

THX-RNMarsh
 
@Demian -- keeping the JFET channel modulation low requires very low signal levels at the drain. Some N/P complementary circuit to linearize the total channel resistance and get rid of the shunt resistor might be more useful. I don't trust simulation for this stuff and just haven't had the time to get back around to the oscillators I have in house suitable for messing with.

@Richard -- I left the 339 as-was except for the trimmer on the reduced R50 (or was it R51? -- which ever was most accessible) for THD -- did not want to dig into it further until I was done with the mods to the Heath IG-18s -- then I just drifted away from coming back to see about improving the notch filter. Perhaps the oscillator section could be even better, as you and David found, by using an LT1468.

Maybe next winter. My HP 3458A could use a cal check at Loveland, which will hurt the account balance, and I've now got several old dogs on the bench waiting for attention. Meanwhile, I'm hoping that by mid-winter, David will have ironed out the remaining wrinkles in his oscillator and I can devote some serious attention to that.
 
Disabled Account
Joined 2012
Thanks Dick as you have a lot of experience and your site is great. I too am looking forward to Davids work as I know it will be something very good. Its all a learning experience for me and just what the levels are and why has been educational for me.
Now that I have several osc, stock and modified, and analyzers available as well, I am pretty satisfied and will be able to move back into amplifiers -- their design and test/measurement.

BTW - I feel the same as Demian and you regarding SIM of osc circuits.

Thx-Dick Marsh
 
Last edited:
A common topology for oscillators is to include a jFET (Q2) control element. With these topologies, THD can be greatly reduced (esp. 2H) by adding 50% feedback from gate to drain. But, just placing twoequal value resistors as shown in the attachment (R50,51) doesnt reduce all the 2nd harmonic possible to reduce. Also, the series R50 with the gate creates some weak 3rd at >10Khz.

Further to THD reduction, just setting the gain control pot. (R30) to a +/- 20v p-p spec. doesnt guarantee lowest THD, either.

So, after replacing the oscillator amplifier (U1) with a lower noise and lower THD opamp (such as, the LT1486) I replace R50,51 and R30 with multi-turn trimmers and adjust them for lowest THD. First the R30 for a THD null. Then R51 for 2H at 100-1KHz and then R50 for lowest H3 at 10Khz. Keep doing this as they all interact and I get an additional 15-20dB reduction in distortion.

This has also been done on certain K-Hite osc models which use this topology with equallly good reductions. Note also that U2C output Z must be low and constant at all freq to get best results over large freq range; U2C output Z is in series with R50 and if its Zo changes, so will the harmonic null tuning.

Thx-RNMarsh


View attachment 357813

Hi Richard,

These are good points. I use the feedback technique for the JFET in the oscillator for my THD Analyzer construction project which appeared in Audio in 1981. It can be found at CordellAudio.com - Home. I employed a fairly interesting way to obtain the drain voltage/2 feedback that you might find interesting. In an upgrade of the analyzer about 3 years ago I also discovered that the 50% feedback of drain signal to the JFET gate is not necessarily optimum. I put in a trimmer to adjust that value and also saw significant improvement. I also use the same JFET control circuit for the amplitude and frequency control circuits in the auto-null of the analyzer.

The whole oscillator+analyzer does about 0.0004% at 1kHz for THD+N. I think the oscillator by itself does about 0.0002% at 1kHz. The analyzer-oscillator combination does about 0.0008% THD+N at 20kHz with a 200kHz measurement bandwidth.

Not bad for 30 years ago, but it could still probably be improved based on some of the results that have been achieved here in this excellent thread. The whole thing used 5534 op amps in the signal path.

Cheers,
Bob
 
Hi Richard,

These are good points. I use the feedback technique for the JFET in the oscillator for my THD Analyzer construction project which appeared in Audio in 1981. It can be found at CordellAudio.com - Home. I employed a fairly interesting way to obtain the drain voltage/2 feedback that you might find interesting. In an upgrade of the analyzer about 3 years ago I also discovered that the 50% feedback of drain signal to the JFET gate is not necessarily optimum. I put in a trimmer to adjust that value and also saw significant improvement. I also use the same JFET control circuit for the amplitude and frequency control circuits in the auto-null of the analyzer.

The whole oscillator+analyzer does about 0.0004% at 1kHz for THD+N. I think the oscillator by itself does about 0.0002% at 1kHz. The analyzer-oscillator combination does about 0.0008% THD+N at 20kHz with a 200kHz measurement bandwidth.

Not bad for 30 years ago, but it could still probably be improved based on some of the results that have been achieved here in this excellent thread. The whole thing used 5534 op amps in the signal path.

Cheers,
Bob

hi bob,

just substituting for the 5534 with a 1468 should improve things alot.
 
Bob is understating the performance of his SVO. With newer opamps, I have achieved THD below at about 0.00006% at 1kHz with Bob's circuit -- with AGC settings that optimize for both 1kHz and 10kHz performance. Optimized only for 1kHz, the THD is around 0.00003%. Given that this is a tunable wide-band oscillator, this is very good performance. See the plot at the bottom of this page -- IG-18 #2, the BIG-18. Note that you need to add three zeros to the ARTA calculated values, due to the 60dB attenuation of the active Twin-T filter ahead of the analyzer.
 
Last edited:
Disabled Account
Joined 2012
THD (+N)

Bob -- Those are exellent numbers by any standards. I did read it when published in 1981. And a couple times more recently and just now to refresh my mind. It has passed the test of time and goes down as one of the greats. A legend.

Until a few months ago, your design was the only THD schematic plus the IG-18 circuitry that I ever saw. I would have built your design but I just could not get my head around the construction of it. Especially the chassis and all.

When I could afford something, i bought an A-P model One dual domain. That served me well until I could not measure the distortion of a simple headphone amp. So, I needed something better whithout breaking the bank again. But, still I didnt want to build from scratch. I'm just not that good with my hands. So, using low cost used commercial products and updating them got me where I am today.

The THD+N of this second 339A runs .0008% (400-80KHz filtering) to 10KHz. The first 339A I was able to get below the noise level of the AD725 analyser but over a limited frequency range [ estimated to be -160dB). One additional area that helps me get so low is the coupling caps. I changed them all to bipolar type, placed back-to-back of highest quality and 105C rating.

The K-H 4400A uses similar topology and my notes say the 'tuned' THD is .0001% to .00025% (1-10KHz). However, the newer K-H 4402B departs for a better control circuitry and I can get THD of .00005 and .0001% from it. They both used the same osc opamp.
[I didnt record the THD+N on them.] I have tried many opamps and the 1468 always works best.

More elaborate commercial brands/models are very good stock when you can find them on the used market.... in the $2k range. They are much, much better than the 339A right out of the box. For example, the VP-7722A has its internal source and analyzer combined THD+N of .0003-7% and THD only of .0001% combined. Or -120dB. I am sure they can be 'tuned' as well using techniques similar to the others.

Then we get to Victor's oscillators... a great value and performance but osc only. I havent found the FFT cards to be accurate enough at low harmonic levels compared to commercial test equipment. And, we still need to try transient recorders on power amps etc.

THX RNMarsh
 
Last edited:
Bob -- Those are exellent numbers by any standards. I did read it when published in 1981. And a couple times more recently and just now to refresh my mind. It has passed the test of time and goes down as one of the greats. A legend.

Until a few months ago, your design was the only THD schematic plus the IG-18 circuitry that I ever saw. I would have built your design but I just could not get my head around the construction of it. Especially the chassis and all.

When I could afford something, i bought an A-P model One dual domain. That served me well until I could not measure the distortion of a simple headphone amp. So, I needed something better whithout breaking the bank again. But, still I didnt want to build from scratch. I'm just not that good with my hands. So, using low cost used commercial products and updating them got me where I am today.

The THD+N of this second 339A runs .0008% (400-80KHz filtering) to 10KHz. The first 339A I was able to get below the noise level of the AD725 analyser but over a limited frequency range [ estimated to be -160dB). One additional area that helps me get so low is the coupling caps. I changed them all to bipolar type, placed back-to-back of highest quality and 105C rating.

The K-H 4400A uses similar topology and my notes say the 'tuned' THD is .0001% to .00025% (1-10KHz). However, the newer K-H 4402B departs for a better control circuitry and I can get THD of .00005 and .0001% from it. They both used the same osc opamp.
[I didnt record the THD+N on them.] I have tried many opamps and the 1468 always works best.

More elaborate commercial brands/models are very good stock when you can find them on the used market.... in the $2k range. They are much, much better than the 339A right out of the box. For example, the VP-7722A has its internal source and analyzer combined THD+N of .0003-7% and THD only of .0001% combined. Or -120dB. I am sure they can be 'tuned' as well using techniques similar to the others.

Then we get to Victor's oscillators... a great value and performance but osc only. I havent found the FFT cards to be accurate enough at low harmonic levels compared to commercial test equipment. And, we still need to try transient recorders on power amps etc.

THX RNMarsh

Hi Richard,

Thanks for your kind words.

This is a great thread, and you deserve much of the credit. My kind of thread. Hard-core design stuff that everybody can learn from - and not a lot of arguing and grand-standing.

I also have a 339A that I picked up for free a couple years ago. I have not been inside of it, but its performance as-is has been a bit disappointing (maybe you get what you pay for). Anyway, this thread has encouraged me to get in there and tweak it when I get some time.

I owe much of my SVO inspiration to Bruce Hofer, an old friend of mine from when he was at Tek designing TM500 modules, including their oscillator/THD analyzer modules. His AES paper on the Tek SVO was what got me going. Of course, he and Rich Cabot subsequently left Tek and went on to found AP.

Cheers,
Bob
 
Disabled Account
Joined 2012
The ShibaSoku combo is at least as good as the best Audio Precision. Here is the combined THD+N at 2v and 1KHz.
But with 100KHz BW. Combined result is: -114dB.
AP gives typical (22KHz BW) individual numbers of .... -115 plus -115dB = ? combined. Maybe -112dB? Plus 1uV each (noise floor).

-RNM

725 THD+N.jpg
 
Last edited:
Disabled Account
Joined 2012
yes, there are some small phase shift going on which cause the distortion null to shift away. It can be tried with a variable C trimmer at the appropriate place while watching the thd meter.

THX-RNMarsh

Take a look at the 339 oscillator schematic I put up, you will see a cap (C47) across R50. This is probably for that HF harmonic reduction function. I will put a variable C in place of C47 and see what I can do with it to reduce THD at higher freqs.

THx-RNMarsh
 
Last edited:
Note that the HP 239 and 339 oscillators deliver their excellent performance at 6.3VRMS output from the oscillator into a medium-Z load; and only somewhat worse at 3.16VRMS into 600 ohms.

I have Bob's SVO set to deliver 1.6VRMS from the output of the second integrator, and use a gain stage to get to nearly 10VRMS with 15V supplies. My measurements reported recently above have been done at 6.3VRMS output for direct comparison to the HP units.
 
Disabled Account
Joined 2012
Bob was using THD+N and so was I when comparing against Audio Precision 2722 as that is how A-P spec their equipment. So, the numbers you have from Bob's design... are they THD+N or THD (using FFT)? You indicated THD.

Does anyone make PCB for Bob's design available today?

Thx-RNMarsh
 
Note that the HP 239 and 339 oscillators deliver their excellent performance at 6.3VRMS output from the oscillator into a medium-Z load; and only somewhat worse at 3.16VRMS into 600 ohms.

I have Bob's SVO set to deliver 1.6VRMS from the output of the second integrator, and use a gain stage to get to nearly 10VRMS with 15V supplies. My measurements reported recently above have been done at 6.3VRMS output for direct comparison to the HP units.

Wow. I'm doing well with my SVO at 3Vrms then. If I ran it at that level the disto would be deep in the noise.