[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [AUDITORY] Differences between RETSPLs for different transducers

Hi Ben,
This is a very interesting question and very important indeed. Let's go step by step so I can share my understanding.
(1) RETSPLs are the result of testing the hearing thresholds of a group of young typical hearing listeners (usually 25 people under 25 years old) with specific transducers.
(2) These transducers have been previously calibrated in dB Sound Pressure Level (SPL) using an ear simulator or just a microphone in the case of free field or diffuse field.
(3) The specification of the RETSPL is then the averaged hearing thresholds corresponding to the hearing tests.
The RETSPL is then the deviation required to play the stimulus in the equivalent dB HL (hearing level). Applying that deviation to the transducers calibrated in dB SPL (i.e. with a flat response using the ear simulator) we assume that we are then playing dB HL.
Does anyone know of any analyses that explain the differences in the RETSPLs for different transducers obtained with ear simulators (rather than just standard couplers)?
I do not know, but I would say that it is not the standard couplers but it is probably related to the losses (leakage). If you look at the RETSPL of the TDH39 (supra-aural), you can clearly see that at low frequency you need 15 dB more than with the HDA-200 (circumaural). To calibrate these two in transducers you need different couplers, but they are also the couplers used in the first place.
On the other side of the coin, we have the Free-field RETSPL which show a characteristic dip at 3-4kHz. That is probably because the calibration of the transducers (2) for the test (1) was done with a free-field microphone, so the hearing threshold results reflect the resonance of the ear canal.
I don't know about any study that compares hearing thresholds using different transducers after applying the RETSPL. That would be very interesting to see, especially for people with hearing loss (symmetric and asymmetric). 

Best Regards

Raul Sanchez-Lopez
Hearing Researcher | Audio Engineer | Technical Audiologist
On Fri, 19 Jul 2024 at 16:20, Ben Lineton <000002db5b7c3f83-dmarc-request@xxxxxxxxxxxxxxx> wrote: 
Dear list,
Does anyone know of any analyses that explain the differences in the RETSPLs for different transducers obtained with ear simulators (rather than just standard couplers)? If we assume a person’s hearing threshold is determined by the SPL at the eardrum roughly independent of tranducer and further assume that an ear simulator achieves a transfer impedance close to that of the average adult ear, then we would expect these RETSPLs to be very similar (and equal to measurements of the minimum audible pressure under headphones).
The attached figure shows the RETSPLs for six transducers measured with three ear simulators: the IEC318-4 (was IEC711) for insert phones, the IEC318-1 for TDH49 supra-aurals, and the IEC318-1 with flat plate for the circumaurals (HDA200,250,300). The figure below show a large differences between RETSPLs (around 20 dB) at 125 Hz and 3 kHz. (Also shown for comparison is the freefield RTSPL measured at the head location plus the head-related transfer function to give the eardrum SPL plus a 2 dB correction for binaural listening; see e.g. Brian Moore, Intro to Psychology of Hearing for explanation).
Explanations for these differences that I’m aware of or seem plausible are these:
1. Detection thresholds are not just determined by the SPL at the eardrum, but are affected by physiological noise which depends to some extent on the ear canal termination of different headphones (hence violating the first assumption above).
2. The acoustic transfer impedance of the ear simulators differ systematically from that of the notional target population (hence violating the second assumption above).
3. Possibly effect of the point acoustic impedance which may differ between the average real ear and the ear simulator. 

4. Sampling errors in determining the RETSPLs.
I wondered if there other factors that I haven’t thought of, and is there any analysis quantifying these different factors (and their frequency dependence)? 

Thanks

Ben









Ben Lineton

University of Southampton
--
Raul Sanchez-Lopez
Researcher