Bob,
In your article on distortion magnifier, there are some very detailed descriptions but there doesn't seem to be a schematics. Would there be one that you can post ?
I guess most of the gain adjusts and sums and differences I can handle. The phase adjust I find a bit tricky. The only circuit I know is the all pass phase adjust as on the Linkwitz website, but that only works at one particular frequency. So do we need to gang switch components (Rs and Cs) to suit the test frequency ? That would also rule out using sweep sine as test signal ?
Thanks in advance,
Patrick
In your article on distortion magnifier, there are some very detailed descriptions but there doesn't seem to be a schematics. Would there be one that you can post ?
I guess most of the gain adjusts and sums and differences I can handle. The phase adjust I find a bit tricky. The only circuit I know is the all pass phase adjust as on the Linkwitz website, but that only works at one particular frequency. So do we need to gang switch components (Rs and Cs) to suit the test frequency ? That would also rule out using sweep sine as test signal ?
Thanks in advance,
Patrick
Hi Patrick,
I'm sorry I'm slow in responding, but I have been getting hammered at work and at home this week.
The schematics for the DM are hand-drawn, several pages. The DM was built on perf board. Send me email if you would like a copy. Ditto for the variable baffle step circuit I used in the Athenas.
You don't need an all-pass or other fancy form of phase shiter to get the null. I just use a single pole LPF. The reason this works is that the 3 dB point of most power amps is above 200 kHz, and the roll-off is often initially close to 6 dB/octave. A decade below, at 20 kHz, the effect of the amp rolloff looks a lot like a constant delay. So does that of the single-pole rolloff I use to match it. It works quite well, but it needs minor adjustment for lower test frequencies. There is also a little bit of interaction with the amplitude matching control. Both have coarse and fine pots.
If one wanted to test at rather low frequencies, some phase adjustment would have to be included at the low-frequency end to account for the AC coupling roll-off in the amplifier. I have not implemented this.
Cheers,
Bob
A similar idea by Wurcer & Jung.
http://www.analog.com/static/imported-files/application_notes/5866763300941AN245.pdf
Take a look at Fig. 4
http://www.analog.com/static/imported-files/application_notes/5866763300941AN245.pdf
Take a look at Fig. 4
Cordell THD Analyzer
I decided to build the PCB making some changes, such as double-sided.
Question:
Using the component side as the ground plane, would improve or worsen things, parasitic capacitance, self-oscillation or similar ?
Thanks and regards
Giuliano
I decided to build the PCB making some changes, such as double-sided.
Question:
Using the component side as the ground plane, would improve or worsen things, parasitic capacitance, self-oscillation or similar ?
Thanks and regards
Giuliano
Here's a photograph of the Source Card for a Boonton 1120 THD% Analyzer -- it's certainly in the same league as the Tektronix SG-505 -- you can see that they use a ground plane on the component layer, but are careful in laying out the leads from the JFET switches such that they aren't surrounded by the copper pour.
The Boonton uses NE5532's. A lot of other chips are "unobtanium".
Most important thing is to keep noise to as low a level as humanly possible. Each card of the Boonton is in it's own little compartment. If I were to build the Cordell I would probably use the Jung-Didden Super Regulators and pay careful attention to placement of the power transformer, rectifier diodes etc. In a lot of the Krohn Hite and HP equipment I have played with over the years they go to great lengths to keep the power supply eminations away and on its own. The one shortcoming of the Boonton is that the original filter caps weren't properly rated for the heat generated by the equipment and the oscillator will let some a.c. line frequency past.
The Boonton uses NE5532's. A lot of other chips are "unobtanium".
Most important thing is to keep noise to as low a level as humanly possible. Each card of the Boonton is in it's own little compartment. If I were to build the Cordell I would probably use the Jung-Didden Super Regulators and pay careful attention to placement of the power transformer, rectifier diodes etc. In a lot of the Krohn Hite and HP equipment I have played with over the years they go to great lengths to keep the power supply eminations away and on its own. The one shortcoming of the Boonton is that the original filter caps weren't properly rated for the heat generated by the equipment and the oscillator will let some a.c. line frequency past.
An externally hosted image should be here but it was not working when we last tested it.
Thanks jackinnj,
very interesting that the Boonton has used JFET for resistors and caps switch .
I am tempted to use the analog-switches type ADG408 or similar.
I think that will put the various stages into metal box.
Again LME49710 instead of NE5534 and LM318.
The power supply will be the Jung super regulator with LT1028 (already built for my PRE Elektor MC).
We'll see ....
very interesting that the Boonton has used JFET for resistors and caps switch .
I am tempted to use the analog-switches type ADG408 or similar.
I think that will put the various stages into metal box.
Again LME49710 instead of NE5534 and LM318.
The power supply will be the Jung super regulator with LT1028 (already built for my PRE Elektor MC).
We'll see ....
I would say that it can only improve it.
Did you put the analyzer all on one PC, or did you keep it on three?
Will you use those great National op amps instead of the 5534's I used?
You are brave for tackling this challenging project.
Best,
Bob
Per Anders sent me some of his regulator boards to test and I am thus an enthusiastic supporter. They've got a smaller footprint than the Jung-Didden boards. OTOH, the J-D boards have been vetted by those far more expert in these matters than me.
Those Signetics opamps are pretty darned good,fwiw. If you could use reed relays they would probably be even better than the semi switches.
Those Signetics opamps are pretty darned good,fwiw. If you could use reed relays they would probably be even better than the semi switches.
These are all good points. I did use a small toroidal transformer and power supply all in a metal box, but doing it today I would be tempted to use an external wall wart with an external first-stage power supply in-line to the analyzer, keeping all the supply hum and ripple out of the analyzer. I did this with my preamp and it worked well. The wall wart supplied AC, that was half-wave rectified into pos and neg raw in the external box, where it was also pre-regulated to something like +/- 18 or 20V. So regulated DC was going into the preamp. Inside the preamp it was re-regulated down to +/-15V.
Note that re-regulation also lets you re-reference the ground to the local ground, which can help with hum and noise. I just used LM317/337 all around.
Cheers,
Bob
Hi all,
my realization provides a new set of PCB (not know yet exactly how many).
Surely there will be the PCB for: Signal source; Voltage-controlled bandpass filter; Product aplifier and level sets; Amplitude and frequency detector; Out filter; meter.
Already think to include in the shield box attenuator, IC9 (input gain = 3 chip)
PCBs will have connectors for the portion of the resistors and capacitors switches, so I can try different possibilities between mechanical and electronic switch. (My goal is to automate the instrument with a microcontroller)
Yes, I will use the LME49710.
Thanks for the "brave" Bob (but often the brave "die" ! ).
I could also use, as power supply, the Q-MINI (DIYZONE) because it is small and I have several PCBs.
The reed relays are among those most likely that I will use!
Maybe a switching regulator could solve the problems of noise?
Thanks for your valuable advice.
Giuliano
my realization provides a new set of PCB (not know yet exactly how many).
Surely there will be the PCB for: Signal source; Voltage-controlled bandpass filter; Product aplifier and level sets; Amplitude and frequency detector; Out filter; meter.
Already think to include in the shield box attenuator, IC9 (input gain = 3 chip)
PCBs will have connectors for the portion of the resistors and capacitors switches, so I can try different possibilities between mechanical and electronic switch. (My goal is to automate the instrument with a microcontroller)
Yes, I will use the LME49710.
Thanks for the "brave" Bob (but often the brave "die" ! ).
I could also use, as power supply, the Q-MINI (DIYZONE) because it is small and I have several PCBs.
The reed relays are among those most likely that I will use!
Maybe a switching regulator could solve the problems of noise?
Thanks for your valuable advice.
Giuliano
I have always shied away from the use of electronic switching devices or variable gain devices in the state variable filters for fear of distortion, but admit that I have not agressively tried it. It would be so nice to eliminate those big old rotary switches and get all the tuning parts down on the board. I did think that if I ever did it again I would probably use trees of relays to do the switching. BTW, I had to work pretty hard to keep down the distortion in the JFET AGC circuit I used in the state variable oscillator.
Another thing that I would do differently is the AGC rectifier in the oscillator. I'd probably go for a four-phase rectifier, rather than the full-wave approach, given that the state variable oscillator naturally provides the quadrature signals. I'd probably also consider true-rms detector ICs to make the rectifier as well. I think AGC rectifier ripple is one of the remaining small sources of distortion in the oscillator.
BTW, if you really want to go crazy you can do an 8-phase rectifier because you can mix the quadrature phases to synthesize a set of four phases that are shifted by 45 degrees.
Cheers,
Bob
Hi,
Another way to do the agc is to use the outputs from the statevariable oscilator as sin and cos and use the old trig identity cos^2 + sin^2 = 1. Then, in theory, you
don't have to filter the signal.
There is a a Philips res rep from 1974 ( in german ) by Meyer-Ebrecht describing this.
/örjan
Another way to do the agc is to use the outputs from the statevariable oscilator as sin and cos and use the old trig identity cos^2 + sin^2 = 1. Then, in theory, you
don't have to filter the signal.
There is a a Philips res rep from 1974 ( in german ) by Meyer-Ebrecht describing this.
/örjan
Last edited:
I sold a Krohn-Hite signal generator on EBay a few months ago -- it was a curiousity item for me, but the THD% never got down to that of the Tektronix SG505, not even remotely. It was a sea of relays inside the box. My AP generator just clacks away all day like an old telephone exchange.
The Texas Instruments programmable volume controls are really, really quiet.
Put "Meyer-Ebrecht +Quadrature" in google scholar -- now I have to get my wife to order the articles.
The Texas Instruments programmable volume controls are really, really quiet.
Put "Meyer-Ebrecht +Quadrature" in google scholar -- now I have to get my wife to order the articles.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.