I used to fix 339A back in the 80's the knobs were cracked even back then. PCB were made in Loveland (PRCD), they liked gold electroplate for some reason or another, I guess because everything else was done that way. Or they did not need solder mask if they gold plated the copper. Probably were all hand solder as well. The 333 series was much better mechanical design from what i can remember. Yes Mr. Marsh you are resurecting a boat anchor, good luck with your efforts.
Cheers
Rick
Cheers
Rick
guts and gore -
Its good we have it noted here for the record for others thinking of doing similar with 339A. Thx-RNMarsh
I am looking for a nice clean sine wave source of variable freq - that i can then run thru a passive filter notch. I plan to use a high quality ADC/DAC system to measure the thd and FFT it in computer. Before that I had 2 fixed freq oscillators from Victor via eBAY. I bought a couple 339A as junk/parts units to cobble into one working unit... but, we shall see. trouble shooting it is easy but time comsuming and not sure I'll have all the parts needed. i did find some NOS of the HA-opamps for $3 each as back up for repair.
Its good we have it noted here for the record for others thinking of doing similar with 339A. Thx-RNMarsh
Last edited:
Used HP equip from eBay is great stuff. I found these units very affordable.
HP 8656B,8901A,3478A,5316A,436A/8484.
I even have service manuals for them and as you do, buy non-op ones for spare parts. Nothin better than the real deal for spare parts ie. semi's .I know that you can get the ServM from Agilent, have lots but some sch's are hardly readable. I got a spare 8656B from Israel, imagine that $100,$100+ to ship air, still a good deal.
Need a part, I may have it! Good idea Rick, buy junk, remove parts, test,.sell on eBay=$?=Retirement planning. I loved my job at Motorola, fixing this stuff, now i fix PRC low rel. stuff=aweful.
Funny story, bought a HP counter IC on eBay, (Purple side braze w/gold lid/leads) sold as an ornament, pulled from a 5336,works like a charm, I see the same IC being sold on eBay for >100.00
Good idea's about making/sourcing some fine audio oscillators.Thx for that info
Have fun Rick
HP 8656B,8901A,3478A,5316A,436A/8484.
I even have service manuals for them and as you do, buy non-op ones for spare parts. Nothin better than the real deal for spare parts ie. semi's .I know that you can get the ServM from Agilent, have lots but some sch's are hardly readable. I got a spare 8656B from Israel, imagine that $100,$100+ to ship air, still a good deal.
Need a part, I may have it! Good idea Rick, buy junk, remove parts, test,.sell on eBay=$?=Retirement planning. I loved my job at Motorola, fixing this stuff, now i fix PRC low rel. stuff=aweful.
Funny story, bought a HP counter IC on eBay, (Purple side braze w/gold lid/leads) sold as an ornament, pulled from a 5336,works like a charm, I see the same IC being sold on eBay for >100.00
Good idea's about making/sourcing some fine audio oscillators.Thx for that info
Have fun Rick
Richard, I have a 339 off of ebay a couple of years ago and only thing done to it was clean the switches although it could be improved with a recal and the 2nd harmonic mod to the jfet. Victor's boards are a better source I believe. Have you considered installing three or four of them into a 339 hulk and using its atten and switches? They are OK and their power supply is also decent, could be tweaked if desired. Might make a pretty decent oscillator. Dick's links has been posted many times on this thread but it is worth a read to look at a 339 osc and a Cordell SVC osc built into a Heathkit box. #2 IG-18 #2, the BIG-18 #3 IG-18 #3, the IG-339A #4 IG-18 #4, the IG-339B The HP has better shielding, attens, power supply etc so a better base to start from if you dont need continuous tuning and just want to use Victors boards. Larry
Good ideas. I understand from Demian, the 339A oscillator is better than it's analyzer says it is. So, I expect it is less than .001%, stock. However, if I need a really, really super low distortion source, I have the ones from Victor.... together, they should take care of my needs for a long time. With the 339 mods, and contact cleaning, it should do very well.
HP 339 mods
A few tweaks I am going to do to my 339 oscillator section when time available:
Add 2nd harmonic mod to jfet.
Remove fuse from sig out line, HP advised some of them caused thermal distortion at low frequencies.
Determine whether better op amps might lower noise from HP/LP filters as Dick Moore suggested.
Improve oscillator +/- 15V power supply to lower noise output from line and PS IC, maybe try super regs with remote sensing. Current limiting nice though in case a tant cap shorts.
Any other suggestions are welcome although I don't want to attempt a major redo of the circuit, that's what Victor's boards are for.
As far as the analyzer section I don't know what improvements are easily made. The power supply for it can probably be improved but its more difficult than the oscillator due to the increased current requirement. Again suggestions are welcome but the need not as great as I also have a hardware SA and use a sound card FFT.
Larry
A few tweaks I am going to do to my 339 oscillator section when time available:
Add 2nd harmonic mod to jfet.
Remove fuse from sig out line, HP advised some of them caused thermal distortion at low frequencies.
Determine whether better op amps might lower noise from HP/LP filters as Dick Moore suggested.
Improve oscillator +/- 15V power supply to lower noise output from line and PS IC, maybe try super regs with remote sensing. Current limiting nice though in case a tant cap shorts.
Any other suggestions are welcome although I don't want to attempt a major redo of the circuit, that's what Victor's boards are for.
As far as the analyzer section I don't know what improvements are easily made. The power supply for it can probably be improved but its more difficult than the oscillator due to the increased current requirement. Again suggestions are welcome but the need not as great as I also have a hardware SA and use a sound card FFT.
Larry
If the analyzer section is well behaved (doesn't add a lot of its own distortion) its a great front end for a PC sound card. Almost infinite protected dynamic range and the notch extends the distortion floor 20-60 dB depending. A 16 bit sound card, or an 8 bit digital scope adapter even, on the monitor output will show virtually all there is to see.
... and get the auto-nulling which i like better than passive manual. Passive bridged-T might be the last word in measuring lowest thd. But, this comes in a close second and hassel free and a lot more affordable for occassional use.
I plan to do some of the tricks mentioned -- like pot for adjusting null on second harmonic. Its just too easy not to do. Thx-RNMarsh
Last edited:
Nickel ated
They probably are nickel... dont know what the base metal is. I will consider soldering the ones that are off the oscillator pcb to other pcb and thru to output. The rest of the contacts (analyzer) can stay as they are.
They probably are nickel... dont know what the base metal is. I will consider soldering the ones that are off the oscillator pcb to other pcb and thru to output. The rest of the contacts (analyzer) can stay as they are.
FFT
I have several FFT options to use... my TEK scope with FFT option to 300MHz, an old, slow analog FFT -HP 3580A (used with my old Sound-Tech but gave that to local college years ago) or a new 24/192kb ADC/DAC and PC/FFT software. Maybe even an iPAD with an FFT application. All of which will get me way below the sine wave source.
My main concern was first a very low thd ocillator source and found Victor's ultra low thd but single freq. on eBAY. But, then I wanted a more flexible oscillator with much less distortion than .001% rather than a collection of many single freq ones. Enter HP339A. -Thx RNM
I have several FFT options to use... my TEK scope with FFT option to 300MHz, an old, slow analog FFT -HP 3580A (used with my old Sound-Tech but gave that to local college years ago) or a new 24/192kb ADC/DAC and PC/FFT software. Maybe even an iPAD with an FFT application. All of which will get me way below the sine wave source.
My main concern was first a very low thd ocillator source and found Victor's ultra low thd but single freq. on eBAY. But, then I wanted a more flexible oscillator with much less distortion than .001% rather than a collection of many single freq ones. Enter HP339A. -Thx RNM
My HP 339A has significant 2nd H spike from the notch circuit, which is what sets the min. THD floor for the unit to 10ppm (-100dB) at 1kHz. The notch null was around -115dB as I recall. I need to make some plots for Davada, so I'll post a couple when I take them. I have some old ones somewhere on this computer, but good luck finding any after several years....
Hopeful thd
That is very hopeful news. The plots will be very helpful. Looking forward to them. Thx-RNMarsh
That is very hopeful news. The plots will be very helpful. Looking forward to them. Thx-RNMarsh
399a
Something I'm considering doing is tapping the output of the notch filter in the 339a.
This will bypass the auto set level in the analyzer which shoves the noise floor and distortion way up on the monitor output. Maybe add a connector somewhere on the rear of the chassis or a switch to route the monitor output there.
The auto set level takes the input level to the notch filter and puts it through a RMS convertor. The output of the RMS detector is compared to a reference and dual matched LDR is used to normalize the gain of the notch filter post amplifier to 1Vrms FS.
If the analyzer is set to a sensitivity of -80dBV FS, the notch is -115dB and the THD+N is near the notch the auto set level is applying a lot gain to normalize this to 1Vrms FS.
Furthermore this gain from the auto set level is provided with an LDR operating in a non ideal part of it's range which means a lot of distortion is added by the post amplifier as well as noise gain.
Additionally, it would be useful to tap the output of the auto set level's RMS detector to provide a level for scaling and this level is the input level to the notch filter.
These are simple mods that shorten the signal chain and provide a much more ideal sound card interface to the 339a.
I was able to get a notch to-125dB by replacing the notch filter op amp with a jfet input type.
Something I'm considering doing is tapping the output of the notch filter in the 339a.
This will bypass the auto set level in the analyzer which shoves the noise floor and distortion way up on the monitor output. Maybe add a connector somewhere on the rear of the chassis or a switch to route the monitor output there.
The auto set level takes the input level to the notch filter and puts it through a RMS convertor. The output of the RMS detector is compared to a reference and dual matched LDR is used to normalize the gain of the notch filter post amplifier to 1Vrms FS.
If the analyzer is set to a sensitivity of -80dBV FS, the notch is -115dB and the THD+N is near the notch the auto set level is applying a lot gain to normalize this to 1Vrms FS.
Furthermore this gain from the auto set level is provided with an LDR operating in a non ideal part of it's range which means a lot of distortion is added by the post amplifier as well as noise gain.
Additionally, it would be useful to tap the output of the auto set level's RMS detector to provide a level for scaling and this level is the input level to the notch filter.
These are simple mods that shorten the signal chain and provide a much more ideal sound card interface to the 339a.
I was able to get a notch to-125dB by replacing the notch filter op amp with a jfet input type.
399a
Something I'm considering doing is tapping the output of the notch filter in the 339a.
This will bypass the auto set level in the analyzer which shoves the noise floor and distortion way up on the monitor output. Maybe add a connector somewhere on the rear of the chassis or a switch to route the monitor output there.
The auto set level takes the input level to the notch filter and puts it through a RMS convertor. The output of the RMS detector is compared to a reference and dual matched LDR is used to normalize the gain of the notch filter post amplifier to 1Vrms FS.
If the analyzer is set to a sensitivity of -80dBV FS, the notch is -115dB and the THD+N is near the notch the auto set level is applying a lot gain to normalize this to 1Vrms FS.
Furthermore this gain from the auto set level is provided with an LDR operating in a non ideal part of it's range which means a lot of distortion is added by the post amplifier as well as noise gain.
Additionally, it would be useful to tap the output of the auto set level's RMS detector to provide a level for scaling and this level is the input level to the notch filter.
These are simple mods that shorten the signal chain and provide a much more ideal sound card interface to the 339a.
I was able to get a notch to-125dB by replacing the notch filter op amp with a jfet input type.
Something I'm considering doing is tapping the output of the notch filter in the 339a.
This will bypass the auto set level in the analyzer which shoves the noise floor and distortion way up on the monitor output. Maybe add a connector somewhere on the rear of the chassis or a switch to route the monitor output there.
The auto set level takes the input level to the notch filter and puts it through a RMS convertor. The output of the RMS detector is compared to a reference and dual matched LDR is used to normalize the gain of the notch filter post amplifier to 1Vrms FS.
If the analyzer is set to a sensitivity of -80dBV FS, the notch is -115dB and the THD+N is near the notch the auto set level is applying a lot gain to normalize this to 1Vrms FS.
Furthermore this gain from the auto set level is provided with an LDR operating in a non ideal part of it's range which means a lot of distortion is added by the post amplifier as well as noise gain.
Additionally, it would be useful to tap the output of the auto set level's RMS detector to provide a level for scaling and this level is the input level to the notch filter.
These are simple mods that shorten the signal chain and provide a much more ideal sound card interface to the 339a.
I was able to get a notch to-125dB by replacing the notch filter op amp with a jfet input type.
I think you are saying that the auto set level is done with an LDR pair at the input to the notch filter that maintains the level there at 3.16V RMS. The LDR's will be a limiter for the distortion. You could upgrade them to the Silonex equivalent which are supposed to be better or use several in a series/parallel network to reduce the voltage drop across them which will reduce the distortion. The distortion is voltage (or maybe current, I'm not sure) dependent so a smaller signal is lower distortion.
A constant level in the notch is pretty important for the auto-notch circuits to work. If you tap before the level meter attenuator for an external FFT you will have a pretty predictable level/known distortion level on the monitor out. If your FFT has low enough noise and the notch circuit is low noise then you should have all you need.
The FET opamp probably has better common mode rejection than the original which will reduce the internal distortion.
A constant level in the notch is pretty important for the auto-notch circuits to work. If you tap before the level meter attenuator for an external FFT you will have a pretty predictable level/known distortion level on the monitor out. If your FFT has low enough noise and the notch circuit is low noise then you should have all you need.
The FET opamp probably has better common mode rejection than the original which will reduce the internal distortion.
339a mod
Some added notes.
The input range indicator of the 339a forces the user to set the input range to normalize a 3.162Vrms FS.
If anyone wants to explore the above mods I would suggest using resistive padding at the output of the notch filter to bring the FS from 3.162Vrms to 1Vrms. This is in range for a sound card. If padding is used it should match the input impedance of the sound card to avoid reflection in the coax cable from impedance mismatch. This is something that should be observed with instrumentation.
Some added notes.
The input range indicator of the 339a forces the user to set the input range to normalize a 3.162Vrms FS.
If anyone wants to explore the above mods I would suggest using resistive padding at the output of the notch filter to bring the FS from 3.162Vrms to 1Vrms. This is in range for a sound card. If padding is used it should match the input impedance of the sound card to avoid reflection in the coax cable from impedance mismatch. This is something that should be observed with instrumentation.
Last edited:
Hi Demian,
I'm not referring to the LDR in the notch filter. The auto set level is done on the post side of the notch filter. It auto scales the output of the notch filter referenced to the input level into the notch filter. It raises the noise significantly when the input level is below the FS. If the input level is 1Vrms then the auto set level increases the gain a little more than 3 times above unity on the post side of the notch filter. There is a long chain of amplifiers, active filters and attenuators in the path before the meter's RMS detector. The monitor taps the meter's RMS detector input. In level mode the monitor is switched to the 339a input amplifier bypassing the analyzer and the noise in this mode is significantly lower.
The mod would eliminate a large part of the signal chain. The only thing effected by doing this would be the scaling and that we can control.
The noise floor from the monitor output is quite significant. I waiting for Dick's measurements to confirm this as I can't find the email with the screen shot I sent him last February and I'm using a different analyzer now which is still in beta test and I'm not convinced.
It seems the sound card FFT method benefits most by having a minimal signal chain.
The Jfet op amp I used doesn't have as good a CMRR as the HA2 2625 I replaced but does a better job of null. I suspect the input loading on the network is at effect here.
Last edited: